1 @c Copyright (C) 1988,1989,1992,1993,1994,1995,1996,1997,1998,1999,2000,2001,
2 @c 2002, 2003, 2004, 2005, 2006, 2007, 2008, 2009
3 @c Free Software Foundation, Inc.
4 @c This is part of the GCC manual.
5 @c For copying conditions, see the file gcc.texi.
8 @chapter Target Description Macros and Functions
9 @cindex machine description macros
10 @cindex target description macros
11 @cindex macros, target description
12 @cindex @file{tm.h} macros
14 In addition to the file @file{@var{machine}.md}, a machine description
15 includes a C header file conventionally given the name
16 @file{@var{machine}.h} and a C source file named @file{@var{machine}.c}.
17 The header file defines numerous macros that convey the information
18 about the target machine that does not fit into the scheme of the
19 @file{.md} file. The file @file{tm.h} should be a link to
20 @file{@var{machine}.h}. The header file @file{config.h} includes
21 @file{tm.h} and most compiler source files include @file{config.h}. The
22 source file defines a variable @code{targetm}, which is a structure
23 containing pointers to functions and data relating to the target
24 machine. @file{@var{machine}.c} should also contain their definitions,
25 if they are not defined elsewhere in GCC, and other functions called
26 through the macros defined in the @file{.h} file.
29 * Target Structure:: The @code{targetm} variable.
30 * Driver:: Controlling how the driver runs the compilation passes.
31 * Run-time Target:: Defining @samp{-m} options like @option{-m68000} and @option{-m68020}.
32 * Per-Function Data:: Defining data structures for per-function information.
33 * Storage Layout:: Defining sizes and alignments of data.
34 * Type Layout:: Defining sizes and properties of basic user data types.
35 * Registers:: Naming and describing the hardware registers.
36 * Register Classes:: Defining the classes of hardware registers.
37 * Old Constraints:: The old way to define machine-specific constraints.
38 * Stack and Calling:: Defining which way the stack grows and by how much.
39 * Varargs:: Defining the varargs macros.
40 * Trampolines:: Code set up at run time to enter a nested function.
41 * Library Calls:: Controlling how library routines are implicitly called.
42 * Addressing Modes:: Defining addressing modes valid for memory operands.
43 * Anchored Addresses:: Defining how @option{-fsection-anchors} should work.
44 * Condition Code:: Defining how insns update the condition code.
45 * Costs:: Defining relative costs of different operations.
46 * Scheduling:: Adjusting the behavior of the instruction scheduler.
47 * Sections:: Dividing storage into text, data, and other sections.
48 * PIC:: Macros for position independent code.
49 * Assembler Format:: Defining how to write insns and pseudo-ops to output.
50 * Debugging Info:: Defining the format of debugging output.
51 * Floating Point:: Handling floating point for cross-compilers.
52 * Mode Switching:: Insertion of mode-switching instructions.
53 * Target Attributes:: Defining target-specific uses of @code{__attribute__}.
54 * Emulated TLS:: Emulated TLS support.
55 * MIPS Coprocessors:: MIPS coprocessor support and how to customize it.
56 * PCH Target:: Validity checking for precompiled headers.
57 * C++ ABI:: Controlling C++ ABI changes.
58 * Misc:: Everything else.
61 @node Target Structure
62 @section The Global @code{targetm} Variable
64 @cindex target functions
66 @deftypevar {struct gcc_target} targetm
67 The target @file{.c} file must define the global @code{targetm} variable
68 which contains pointers to functions and data relating to the target
69 machine. The variable is declared in @file{target.h};
70 @file{target-def.h} defines the macro @code{TARGET_INITIALIZER} which is
71 used to initialize the variable, and macros for the default initializers
72 for elements of the structure. The @file{.c} file should override those
73 macros for which the default definition is inappropriate. For example:
76 #include "target-def.h"
78 /* @r{Initialize the GCC target structure.} */
80 #undef TARGET_COMP_TYPE_ATTRIBUTES
81 #define TARGET_COMP_TYPE_ATTRIBUTES @var{machine}_comp_type_attributes
83 struct gcc_target targetm = TARGET_INITIALIZER;
87 Where a macro should be defined in the @file{.c} file in this manner to
88 form part of the @code{targetm} structure, it is documented below as a
89 ``Target Hook'' with a prototype. Many macros will change in future
90 from being defined in the @file{.h} file to being part of the
91 @code{targetm} structure.
94 @section Controlling the Compilation Driver, @file{gcc}
96 @cindex controlling the compilation driver
98 @c prevent bad page break with this line
99 You can control the compilation driver.
101 @defmac SWITCH_TAKES_ARG (@var{char})
102 A C expression which determines whether the option @option{-@var{char}}
103 takes arguments. The value should be the number of arguments that
104 option takes--zero, for many options.
106 By default, this macro is defined as
107 @code{DEFAULT_SWITCH_TAKES_ARG}, which handles the standard options
108 properly. You need not define @code{SWITCH_TAKES_ARG} unless you
109 wish to add additional options which take arguments. Any redefinition
110 should call @code{DEFAULT_SWITCH_TAKES_ARG} and then check for
114 @defmac WORD_SWITCH_TAKES_ARG (@var{name})
115 A C expression which determines whether the option @option{-@var{name}}
116 takes arguments. The value should be the number of arguments that
117 option takes--zero, for many options. This macro rather than
118 @code{SWITCH_TAKES_ARG} is used for multi-character option names.
120 By default, this macro is defined as
121 @code{DEFAULT_WORD_SWITCH_TAKES_ARG}, which handles the standard options
122 properly. You need not define @code{WORD_SWITCH_TAKES_ARG} unless you
123 wish to add additional options which take arguments. Any redefinition
124 should call @code{DEFAULT_WORD_SWITCH_TAKES_ARG} and then check for
128 @defmac SWITCH_CURTAILS_COMPILATION (@var{char})
129 A C expression which determines whether the option @option{-@var{char}}
130 stops compilation before the generation of an executable. The value is
131 boolean, nonzero if the option does stop an executable from being
132 generated, zero otherwise.
134 By default, this macro is defined as
135 @code{DEFAULT_SWITCH_CURTAILS_COMPILATION}, which handles the standard
136 options properly. You need not define
137 @code{SWITCH_CURTAILS_COMPILATION} unless you wish to add additional
138 options which affect the generation of an executable. Any redefinition
139 should call @code{DEFAULT_SWITCH_CURTAILS_COMPILATION} and then check
140 for additional options.
143 @defmac SWITCHES_NEED_SPACES
144 A string-valued C expression which enumerates the options for which
145 the linker needs a space between the option and its argument.
147 If this macro is not defined, the default value is @code{""}.
150 @defmac TARGET_OPTION_TRANSLATE_TABLE
151 If defined, a list of pairs of strings, the first of which is a
152 potential command line target to the @file{gcc} driver program, and the
153 second of which is a space-separated (tabs and other whitespace are not
154 supported) list of options with which to replace the first option. The
155 target defining this list is responsible for assuring that the results
156 are valid. Replacement options may not be the @code{--opt} style, they
157 must be the @code{-opt} style. It is the intention of this macro to
158 provide a mechanism for substitution that affects the multilibs chosen,
159 such as one option that enables many options, some of which select
160 multilibs. Example nonsensical definition, where @option{-malt-abi},
161 @option{-EB}, and @option{-mspoo} cause different multilibs to be chosen:
164 #define TARGET_OPTION_TRANSLATE_TABLE \
165 @{ "-fast", "-march=fast-foo -malt-abi -I/usr/fast-foo" @}, \
166 @{ "-compat", "-EB -malign=4 -mspoo" @}
170 @defmac DRIVER_SELF_SPECS
171 A list of specs for the driver itself. It should be a suitable
172 initializer for an array of strings, with no surrounding braces.
174 The driver applies these specs to its own command line between loading
175 default @file{specs} files (but not command-line specified ones) and
176 choosing the multilib directory or running any subcommands. It
177 applies them in the order given, so each spec can depend on the
178 options added by earlier ones. It is also possible to remove options
179 using @samp{%<@var{option}} in the usual way.
181 This macro can be useful when a port has several interdependent target
182 options. It provides a way of standardizing the command line so
183 that the other specs are easier to write.
185 Do not define this macro if it does not need to do anything.
188 @defmac OPTION_DEFAULT_SPECS
189 A list of specs used to support configure-time default options (i.e.@:
190 @option{--with} options) in the driver. It should be a suitable initializer
191 for an array of structures, each containing two strings, without the
192 outermost pair of surrounding braces.
194 The first item in the pair is the name of the default. This must match
195 the code in @file{config.gcc} for the target. The second item is a spec
196 to apply if a default with this name was specified. The string
197 @samp{%(VALUE)} in the spec will be replaced by the value of the default
198 everywhere it occurs.
200 The driver will apply these specs to its own command line between loading
201 default @file{specs} files and processing @code{DRIVER_SELF_SPECS}, using
202 the same mechanism as @code{DRIVER_SELF_SPECS}.
204 Do not define this macro if it does not need to do anything.
208 A C string constant that tells the GCC driver program options to
209 pass to CPP@. It can also specify how to translate options you
210 give to GCC into options for GCC to pass to the CPP@.
212 Do not define this macro if it does not need to do anything.
215 @defmac CPLUSPLUS_CPP_SPEC
216 This macro is just like @code{CPP_SPEC}, but is used for C++, rather
217 than C@. If you do not define this macro, then the value of
218 @code{CPP_SPEC} (if any) will be used instead.
222 A C string constant that tells the GCC driver program options to
223 pass to @code{cc1}, @code{cc1plus}, @code{f771}, and the other language
225 It can also specify how to translate options you give to GCC into options
226 for GCC to pass to front ends.
228 Do not define this macro if it does not need to do anything.
232 A C string constant that tells the GCC driver program options to
233 pass to @code{cc1plus}. It can also specify how to translate options you
234 give to GCC into options for GCC to pass to the @code{cc1plus}.
236 Do not define this macro if it does not need to do anything.
237 Note that everything defined in CC1_SPEC is already passed to
238 @code{cc1plus} so there is no need to duplicate the contents of
239 CC1_SPEC in CC1PLUS_SPEC@.
243 A C string constant that tells the GCC driver program options to
244 pass to the assembler. It can also specify how to translate options
245 you give to GCC into options for GCC to pass to the assembler.
246 See the file @file{sun3.h} for an example of this.
248 Do not define this macro if it does not need to do anything.
251 @defmac ASM_FINAL_SPEC
252 A C string constant that tells the GCC driver program how to
253 run any programs which cleanup after the normal assembler.
254 Normally, this is not needed. See the file @file{mips.h} for
257 Do not define this macro if it does not need to do anything.
260 @defmac AS_NEEDS_DASH_FOR_PIPED_INPUT
261 Define this macro, with no value, if the driver should give the assembler
262 an argument consisting of a single dash, @option{-}, to instruct it to
263 read from its standard input (which will be a pipe connected to the
264 output of the compiler proper). This argument is given after any
265 @option{-o} option specifying the name of the output file.
267 If you do not define this macro, the assembler is assumed to read its
268 standard input if given no non-option arguments. If your assembler
269 cannot read standard input at all, use a @samp{%@{pipe:%e@}} construct;
270 see @file{mips.h} for instance.
274 A C string constant that tells the GCC driver program options to
275 pass to the linker. It can also specify how to translate options you
276 give to GCC into options for GCC to pass to the linker.
278 Do not define this macro if it does not need to do anything.
282 Another C string constant used much like @code{LINK_SPEC}. The difference
283 between the two is that @code{LIB_SPEC} is used at the end of the
284 command given to the linker.
286 If this macro is not defined, a default is provided that
287 loads the standard C library from the usual place. See @file{gcc.c}.
291 Another C string constant that tells the GCC driver program
292 how and when to place a reference to @file{libgcc.a} into the
293 linker command line. This constant is placed both before and after
294 the value of @code{LIB_SPEC}.
296 If this macro is not defined, the GCC driver provides a default that
297 passes the string @option{-lgcc} to the linker.
300 @defmac REAL_LIBGCC_SPEC
301 By default, if @code{ENABLE_SHARED_LIBGCC} is defined, the
302 @code{LIBGCC_SPEC} is not directly used by the driver program but is
303 instead modified to refer to different versions of @file{libgcc.a}
304 depending on the values of the command line flags @option{-static},
305 @option{-shared}, @option{-static-libgcc}, and @option{-shared-libgcc}. On
306 targets where these modifications are inappropriate, define
307 @code{REAL_LIBGCC_SPEC} instead. @code{REAL_LIBGCC_SPEC} tells the
308 driver how to place a reference to @file{libgcc} on the link command
309 line, but, unlike @code{LIBGCC_SPEC}, it is used unmodified.
312 @defmac USE_LD_AS_NEEDED
313 A macro that controls the modifications to @code{LIBGCC_SPEC}
314 mentioned in @code{REAL_LIBGCC_SPEC}. If nonzero, a spec will be
315 generated that uses --as-needed and the shared libgcc in place of the
316 static exception handler library, when linking without any of
317 @code{-static}, @code{-static-libgcc}, or @code{-shared-libgcc}.
321 If defined, this C string constant is added to @code{LINK_SPEC}.
322 When @code{USE_LD_AS_NEEDED} is zero or undefined, it also affects
323 the modifications to @code{LIBGCC_SPEC} mentioned in
324 @code{REAL_LIBGCC_SPEC}.
327 @defmac STARTFILE_SPEC
328 Another C string constant used much like @code{LINK_SPEC}. The
329 difference between the two is that @code{STARTFILE_SPEC} is used at
330 the very beginning of the command given to the linker.
332 If this macro is not defined, a default is provided that loads the
333 standard C startup file from the usual place. See @file{gcc.c}.
337 Another C string constant used much like @code{LINK_SPEC}. The
338 difference between the two is that @code{ENDFILE_SPEC} is used at
339 the very end of the command given to the linker.
341 Do not define this macro if it does not need to do anything.
344 @defmac THREAD_MODEL_SPEC
345 GCC @code{-v} will print the thread model GCC was configured to use.
346 However, this doesn't work on platforms that are multilibbed on thread
347 models, such as AIX 4.3. On such platforms, define
348 @code{THREAD_MODEL_SPEC} such that it evaluates to a string without
349 blanks that names one of the recognized thread models. @code{%*}, the
350 default value of this macro, will expand to the value of
351 @code{thread_file} set in @file{config.gcc}.
354 @defmac SYSROOT_SUFFIX_SPEC
355 Define this macro to add a suffix to the target sysroot when GCC is
356 configured with a sysroot. This will cause GCC to search for usr/lib,
357 et al, within sysroot+suffix.
360 @defmac SYSROOT_HEADERS_SUFFIX_SPEC
361 Define this macro to add a headers_suffix to the target sysroot when
362 GCC is configured with a sysroot. This will cause GCC to pass the
363 updated sysroot+headers_suffix to CPP, causing it to search for
364 usr/include, et al, within sysroot+headers_suffix.
368 Define this macro to provide additional specifications to put in the
369 @file{specs} file that can be used in various specifications like
372 The definition should be an initializer for an array of structures,
373 containing a string constant, that defines the specification name, and a
374 string constant that provides the specification.
376 Do not define this macro if it does not need to do anything.
378 @code{EXTRA_SPECS} is useful when an architecture contains several
379 related targets, which have various @code{@dots{}_SPECS} which are similar
380 to each other, and the maintainer would like one central place to keep
383 For example, the PowerPC System V.4 targets use @code{EXTRA_SPECS} to
384 define either @code{_CALL_SYSV} when the System V calling sequence is
385 used or @code{_CALL_AIX} when the older AIX-based calling sequence is
388 The @file{config/rs6000/rs6000.h} target file defines:
391 #define EXTRA_SPECS \
392 @{ "cpp_sysv_default", CPP_SYSV_DEFAULT @},
394 #define CPP_SYS_DEFAULT ""
397 The @file{config/rs6000/sysv.h} target file defines:
401 "%@{posix: -D_POSIX_SOURCE @} \
402 %@{mcall-sysv: -D_CALL_SYSV @} \
403 %@{!mcall-sysv: %(cpp_sysv_default) @} \
404 %@{msoft-float: -D_SOFT_FLOAT@} %@{mcpu=403: -D_SOFT_FLOAT@}"
406 #undef CPP_SYSV_DEFAULT
407 #define CPP_SYSV_DEFAULT "-D_CALL_SYSV"
410 while the @file{config/rs6000/eabiaix.h} target file defines
411 @code{CPP_SYSV_DEFAULT} as:
414 #undef CPP_SYSV_DEFAULT
415 #define CPP_SYSV_DEFAULT "-D_CALL_AIX"
419 @defmac LINK_LIBGCC_SPECIAL_1
420 Define this macro if the driver program should find the library
421 @file{libgcc.a}. If you do not define this macro, the driver program will pass
422 the argument @option{-lgcc} to tell the linker to do the search.
425 @defmac LINK_GCC_C_SEQUENCE_SPEC
426 The sequence in which libgcc and libc are specified to the linker.
427 By default this is @code{%G %L %G}.
430 @defmac LINK_COMMAND_SPEC
431 A C string constant giving the complete command line need to execute the
432 linker. When you do this, you will need to update your port each time a
433 change is made to the link command line within @file{gcc.c}. Therefore,
434 define this macro only if you need to completely redefine the command
435 line for invoking the linker and there is no other way to accomplish
436 the effect you need. Overriding this macro may be avoidable by overriding
437 @code{LINK_GCC_C_SEQUENCE_SPEC} instead.
440 @defmac LINK_ELIMINATE_DUPLICATE_LDIRECTORIES
441 A nonzero value causes @command{collect2} to remove duplicate @option{-L@var{directory}} search
442 directories from linking commands. Do not give it a nonzero value if
443 removing duplicate search directories changes the linker's semantics.
446 @defmac MULTILIB_DEFAULTS
447 Define this macro as a C expression for the initializer of an array of
448 string to tell the driver program which options are defaults for this
449 target and thus do not need to be handled specially when using
450 @code{MULTILIB_OPTIONS}.
452 Do not define this macro if @code{MULTILIB_OPTIONS} is not defined in
453 the target makefile fragment or if none of the options listed in
454 @code{MULTILIB_OPTIONS} are set by default.
455 @xref{Target Fragment}.
458 @defmac RELATIVE_PREFIX_NOT_LINKDIR
459 Define this macro to tell @command{gcc} that it should only translate
460 a @option{-B} prefix into a @option{-L} linker option if the prefix
461 indicates an absolute file name.
464 @defmac MD_EXEC_PREFIX
465 If defined, this macro is an additional prefix to try after
466 @code{STANDARD_EXEC_PREFIX}. @code{MD_EXEC_PREFIX} is not searched
467 when the @option{-b} option is used, or the compiler is built as a cross
468 compiler. If you define @code{MD_EXEC_PREFIX}, then be sure to add it
469 to the list of directories used to find the assembler in @file{configure.in}.
472 @defmac STANDARD_STARTFILE_PREFIX
473 Define this macro as a C string constant if you wish to override the
474 standard choice of @code{libdir} as the default prefix to
475 try when searching for startup files such as @file{crt0.o}.
476 @code{STANDARD_STARTFILE_PREFIX} is not searched when the compiler
477 is built as a cross compiler.
480 @defmac STANDARD_STARTFILE_PREFIX_1
481 Define this macro as a C string constant if you wish to override the
482 standard choice of @code{/lib} as a prefix to try after the default prefix
483 when searching for startup files such as @file{crt0.o}.
484 @code{STANDARD_STARTFILE_PREFIX_1} is not searched when the compiler
485 is built as a cross compiler.
488 @defmac STANDARD_STARTFILE_PREFIX_2
489 Define this macro as a C string constant if you wish to override the
490 standard choice of @code{/lib} as yet another prefix to try after the
491 default prefix when searching for startup files such as @file{crt0.o}.
492 @code{STANDARD_STARTFILE_PREFIX_2} is not searched when the compiler
493 is built as a cross compiler.
496 @defmac MD_STARTFILE_PREFIX
497 If defined, this macro supplies an additional prefix to try after the
498 standard prefixes. @code{MD_EXEC_PREFIX} is not searched when the
499 @option{-b} option is used, or when the compiler is built as a cross
503 @defmac MD_STARTFILE_PREFIX_1
504 If defined, this macro supplies yet another prefix to try after the
505 standard prefixes. It is not searched when the @option{-b} option is
506 used, or when the compiler is built as a cross compiler.
509 @defmac INIT_ENVIRONMENT
510 Define this macro as a C string constant if you wish to set environment
511 variables for programs called by the driver, such as the assembler and
512 loader. The driver passes the value of this macro to @code{putenv} to
513 initialize the necessary environment variables.
516 @defmac LOCAL_INCLUDE_DIR
517 Define this macro as a C string constant if you wish to override the
518 standard choice of @file{/usr/local/include} as the default prefix to
519 try when searching for local header files. @code{LOCAL_INCLUDE_DIR}
520 comes before @code{SYSTEM_INCLUDE_DIR} in the search order.
522 Cross compilers do not search either @file{/usr/local/include} or its
526 @defmac MODIFY_TARGET_NAME
527 Define this macro if you wish to define command-line switches that
528 modify the default target name.
530 For each switch, you can include a string to be appended to the first
531 part of the configuration name or a string to be deleted from the
532 configuration name, if present. The definition should be an initializer
533 for an array of structures. Each array element should have three
534 elements: the switch name (a string constant, including the initial
535 dash), one of the enumeration codes @code{ADD} or @code{DELETE} to
536 indicate whether the string should be inserted or deleted, and the string
537 to be inserted or deleted (a string constant).
539 For example, on a machine where @samp{64} at the end of the
540 configuration name denotes a 64-bit target and you want the @option{-32}
541 and @option{-64} switches to select between 32- and 64-bit targets, you would
545 #define MODIFY_TARGET_NAME \
546 @{ @{ "-32", DELETE, "64"@}, \
547 @{"-64", ADD, "64"@}@}
551 @defmac SYSTEM_INCLUDE_DIR
552 Define this macro as a C string constant if you wish to specify a
553 system-specific directory to search for header files before the standard
554 directory. @code{SYSTEM_INCLUDE_DIR} comes before
555 @code{STANDARD_INCLUDE_DIR} in the search order.
557 Cross compilers do not use this macro and do not search the directory
561 @defmac STANDARD_INCLUDE_DIR
562 Define this macro as a C string constant if you wish to override the
563 standard choice of @file{/usr/include} as the default prefix to
564 try when searching for header files.
566 Cross compilers ignore this macro and do not search either
567 @file{/usr/include} or its replacement.
570 @defmac STANDARD_INCLUDE_COMPONENT
571 The ``component'' corresponding to @code{STANDARD_INCLUDE_DIR}.
572 See @code{INCLUDE_DEFAULTS}, below, for the description of components.
573 If you do not define this macro, no component is used.
576 @defmac INCLUDE_DEFAULTS
577 Define this macro if you wish to override the entire default search path
578 for include files. For a native compiler, the default search path
579 usually consists of @code{GCC_INCLUDE_DIR}, @code{LOCAL_INCLUDE_DIR},
580 @code{SYSTEM_INCLUDE_DIR}, @code{GPLUSPLUS_INCLUDE_DIR}, and
581 @code{STANDARD_INCLUDE_DIR}. In addition, @code{GPLUSPLUS_INCLUDE_DIR}
582 and @code{GCC_INCLUDE_DIR} are defined automatically by @file{Makefile},
583 and specify private search areas for GCC@. The directory
584 @code{GPLUSPLUS_INCLUDE_DIR} is used only for C++ programs.
586 The definition should be an initializer for an array of structures.
587 Each array element should have four elements: the directory name (a
588 string constant), the component name (also a string constant), a flag
589 for C++-only directories,
590 and a flag showing that the includes in the directory don't need to be
591 wrapped in @code{extern @samp{C}} when compiling C++. Mark the end of
592 the array with a null element.
594 The component name denotes what GNU package the include file is part of,
595 if any, in all uppercase letters. For example, it might be @samp{GCC}
596 or @samp{BINUTILS}. If the package is part of a vendor-supplied
597 operating system, code the component name as @samp{0}.
599 For example, here is the definition used for VAX/VMS:
602 #define INCLUDE_DEFAULTS \
604 @{ "GNU_GXX_INCLUDE:", "G++", 1, 1@}, \
605 @{ "GNU_CC_INCLUDE:", "GCC", 0, 0@}, \
606 @{ "SYS$SYSROOT:[SYSLIB.]", 0, 0, 0@}, \
613 Here is the order of prefixes tried for exec files:
617 Any prefixes specified by the user with @option{-B}.
620 The environment variable @code{GCC_EXEC_PREFIX} or, if @code{GCC_EXEC_PREFIX}
621 is not set and the compiler has not been installed in the configure-time
622 @var{prefix}, the location in which the compiler has actually been installed.
625 The directories specified by the environment variable @code{COMPILER_PATH}.
628 The macro @code{STANDARD_EXEC_PREFIX}, if the compiler has been installed
629 in the configured-time @var{prefix}.
632 The location @file{/usr/libexec/gcc/}, but only if this is a native compiler.
635 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
638 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
642 Here is the order of prefixes tried for startfiles:
646 Any prefixes specified by the user with @option{-B}.
649 The environment variable @code{GCC_EXEC_PREFIX} or its automatically determined
650 value based on the installed toolchain location.
653 The directories specified by the environment variable @code{LIBRARY_PATH}
654 (or port-specific name; native only, cross compilers do not use this).
657 The macro @code{STANDARD_EXEC_PREFIX}, but only if the toolchain is installed
658 in the configured @var{prefix} or this is a native compiler.
661 The location @file{/usr/lib/gcc/}, but only if this is a native compiler.
664 The macro @code{MD_EXEC_PREFIX}, if defined, but only if this is a native
668 The macro @code{MD_STARTFILE_PREFIX}, if defined, but only if this is a
669 native compiler, or we have a target system root.
672 The macro @code{MD_STARTFILE_PREFIX_1}, if defined, but only if this is a
673 native compiler, or we have a target system root.
676 The macro @code{STANDARD_STARTFILE_PREFIX}, with any sysroot modifications.
677 If this path is relative it will be prefixed by @code{GCC_EXEC_PREFIX} and
678 the machine suffix or @code{STANDARD_EXEC_PREFIX} and the machine suffix.
681 The macro @code{STANDARD_STARTFILE_PREFIX_1}, but only if this is a native
682 compiler, or we have a target system root. The default for this macro is
686 The macro @code{STANDARD_STARTFILE_PREFIX_2}, but only if this is a native
687 compiler, or we have a target system root. The default for this macro is
691 @node Run-time Target
692 @section Run-time Target Specification
693 @cindex run-time target specification
694 @cindex predefined macros
695 @cindex target specifications
697 @c prevent bad page break with this line
698 Here are run-time target specifications.
700 @defmac TARGET_CPU_CPP_BUILTINS ()
701 This function-like macro expands to a block of code that defines
702 built-in preprocessor macros and assertions for the target CPU, using
703 the functions @code{builtin_define}, @code{builtin_define_std} and
704 @code{builtin_assert}. When the front end
705 calls this macro it provides a trailing semicolon, and since it has
706 finished command line option processing your code can use those
709 @code{builtin_assert} takes a string in the form you pass to the
710 command-line option @option{-A}, such as @code{cpu=mips}, and creates
711 the assertion. @code{builtin_define} takes a string in the form
712 accepted by option @option{-D} and unconditionally defines the macro.
714 @code{builtin_define_std} takes a string representing the name of an
715 object-like macro. If it doesn't lie in the user's namespace,
716 @code{builtin_define_std} defines it unconditionally. Otherwise, it
717 defines a version with two leading underscores, and another version
718 with two leading and trailing underscores, and defines the original
719 only if an ISO standard was not requested on the command line. For
720 example, passing @code{unix} defines @code{__unix}, @code{__unix__}
721 and possibly @code{unix}; passing @code{_mips} defines @code{__mips},
722 @code{__mips__} and possibly @code{_mips}, and passing @code{_ABI64}
723 defines only @code{_ABI64}.
725 You can also test for the C dialect being compiled. The variable
726 @code{c_language} is set to one of @code{clk_c}, @code{clk_cplusplus}
727 or @code{clk_objective_c}. Note that if we are preprocessing
728 assembler, this variable will be @code{clk_c} but the function-like
729 macro @code{preprocessing_asm_p()} will return true, so you might want
730 to check for that first. If you need to check for strict ANSI, the
731 variable @code{flag_iso} can be used. The function-like macro
732 @code{preprocessing_trad_p()} can be used to check for traditional
736 @defmac TARGET_OS_CPP_BUILTINS ()
737 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
738 and is used for the target operating system instead.
741 @defmac TARGET_OBJFMT_CPP_BUILTINS ()
742 Similarly to @code{TARGET_CPU_CPP_BUILTINS} but this macro is optional
743 and is used for the target object format. @file{elfos.h} uses this
744 macro to define @code{__ELF__}, so you probably do not need to define
748 @deftypevar {extern int} target_flags
749 This variable is declared in @file{options.h}, which is included before
750 any target-specific headers.
753 @deftypevr {Target Hook} int TARGET_DEFAULT_TARGET_FLAGS
754 This variable specifies the initial value of @code{target_flags}.
755 Its default setting is 0.
758 @cindex optional hardware or system features
759 @cindex features, optional, in system conventions
761 @deftypefn {Target Hook} bool TARGET_HANDLE_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
762 This hook is called whenever the user specifies one of the
763 target-specific options described by the @file{.opt} definition files
764 (@pxref{Options}). It has the opportunity to do some option-specific
765 processing and should return true if the option is valid. The default
766 definition does nothing but return true.
768 @var{code} specifies the @code{OPT_@var{name}} enumeration value
769 associated with the selected option; @var{name} is just a rendering of
770 the option name in which non-alphanumeric characters are replaced by
771 underscores. @var{arg} specifies the string argument and is null if
772 no argument was given. If the option is flagged as a @code{UInteger}
773 (@pxref{Option properties}), @var{value} is the numeric value of the
774 argument. Otherwise @var{value} is 1 if the positive form of the
775 option was used and 0 if the ``no-'' form was.
778 @deftypefn {Target Hook} bool TARGET_HANDLE_C_OPTION (size_t @var{code}, const char *@var{arg}, int @var{value})
779 This target hook is called whenever the user specifies one of the
780 target-specific C language family options described by the @file{.opt}
781 definition files(@pxref{Options}). It has the opportunity to do some
782 option-specific processing and should return true if the option is
783 valid. The default definition does nothing but return false.
785 In general, you should use @code{TARGET_HANDLE_OPTION} to handle
786 options. However, if processing an option requires routines that are
787 only available in the C (and related language) front ends, then you
788 should use @code{TARGET_HANDLE_C_OPTION} instead.
791 @defmac TARGET_VERSION
792 This macro is a C statement to print on @code{stderr} a string
793 describing the particular machine description choice. Every machine
794 description should define @code{TARGET_VERSION}. For example:
798 #define TARGET_VERSION \
799 fprintf (stderr, " (68k, Motorola syntax)");
801 #define TARGET_VERSION \
802 fprintf (stderr, " (68k, MIT syntax)");
807 @defmac OVERRIDE_OPTIONS
808 Sometimes certain combinations of command options do not make sense on
809 a particular target machine. You can define a macro
810 @code{OVERRIDE_OPTIONS} to take account of this. This macro, if
811 defined, is executed once just after all the command options have been
814 Don't use this macro to turn on various extra optimizations for
815 @option{-O}. That is what @code{OPTIMIZATION_OPTIONS} is for.
818 @defmac C_COMMON_OVERRIDE_OPTIONS
819 This is similar to @code{OVERRIDE_OPTIONS} but is only used in the C
820 language frontends (C, Objective-C, C++, Objective-C++) and so can be
821 used to alter option flag variables which only exist in those
825 @defmac OPTIMIZATION_OPTIONS (@var{level}, @var{size})
826 Some machines may desire to change what optimizations are performed for
827 various optimization levels. This macro, if defined, is executed once
828 just after the optimization level is determined and before the remainder
829 of the command options have been parsed. Values set in this macro are
830 used as the default values for the other command line options.
832 @var{level} is the optimization level specified; 2 if @option{-O2} is
833 specified, 1 if @option{-O} is specified, and 0 if neither is specified.
835 @var{size} is nonzero if @option{-Os} is specified and zero otherwise.
837 This macro is run once at program startup and when the optimization
838 options are changed via @code{#pragma GCC optimize} or by using the
839 @code{optimize} attribute.
841 @strong{Do not examine @code{write_symbols} in
842 this macro!} The debugging options are not supposed to alter the
846 @deftypefn {Target Hook} bool TARGET_HELP (void)
847 This hook is called in response to the user invoking
848 @option{--target-help} on the command line. It gives the target a
849 chance to display extra information on the target specific command
850 line options found in its @file{.opt} file.
853 @defmac CAN_DEBUG_WITHOUT_FP
854 Define this macro if debugging can be performed even without a frame
855 pointer. If this macro is defined, GCC will turn on the
856 @option{-fomit-frame-pointer} option whenever @option{-O} is specified.
859 @node Per-Function Data
860 @section Defining data structures for per-function information.
861 @cindex per-function data
862 @cindex data structures
864 If the target needs to store information on a per-function basis, GCC
865 provides a macro and a couple of variables to allow this. Note, just
866 using statics to store the information is a bad idea, since GCC supports
867 nested functions, so you can be halfway through encoding one function
868 when another one comes along.
870 GCC defines a data structure called @code{struct function} which
871 contains all of the data specific to an individual function. This
872 structure contains a field called @code{machine} whose type is
873 @code{struct machine_function *}, which can be used by targets to point
874 to their own specific data.
876 If a target needs per-function specific data it should define the type
877 @code{struct machine_function} and also the macro @code{INIT_EXPANDERS}.
878 This macro should be used to initialize the function pointer
879 @code{init_machine_status}. This pointer is explained below.
881 One typical use of per-function, target specific data is to create an
882 RTX to hold the register containing the function's return address. This
883 RTX can then be used to implement the @code{__builtin_return_address}
884 function, for level 0.
886 Note---earlier implementations of GCC used a single data area to hold
887 all of the per-function information. Thus when processing of a nested
888 function began the old per-function data had to be pushed onto a
889 stack, and when the processing was finished, it had to be popped off the
890 stack. GCC used to provide function pointers called
891 @code{save_machine_status} and @code{restore_machine_status} to handle
892 the saving and restoring of the target specific information. Since the
893 single data area approach is no longer used, these pointers are no
896 @defmac INIT_EXPANDERS
897 Macro called to initialize any target specific information. This macro
898 is called once per function, before generation of any RTL has begun.
899 The intention of this macro is to allow the initialization of the
900 function pointer @code{init_machine_status}.
903 @deftypevar {void (*)(struct function *)} init_machine_status
904 If this function pointer is non-@code{NULL} it will be called once per
905 function, before function compilation starts, in order to allow the
906 target to perform any target specific initialization of the
907 @code{struct function} structure. It is intended that this would be
908 used to initialize the @code{machine} of that structure.
910 @code{struct machine_function} structures are expected to be freed by GC@.
911 Generally, any memory that they reference must be allocated by using
912 @code{ggc_alloc}, including the structure itself.
916 @section Storage Layout
917 @cindex storage layout
919 Note that the definitions of the macros in this table which are sizes or
920 alignments measured in bits do not need to be constant. They can be C
921 expressions that refer to static variables, such as the @code{target_flags}.
922 @xref{Run-time Target}.
924 @defmac BITS_BIG_ENDIAN
925 Define this macro to have the value 1 if the most significant bit in a
926 byte has the lowest number; otherwise define it to have the value zero.
927 This means that bit-field instructions count from the most significant
928 bit. If the machine has no bit-field instructions, then this must still
929 be defined, but it doesn't matter which value it is defined to. This
930 macro need not be a constant.
932 This macro does not affect the way structure fields are packed into
933 bytes or words; that is controlled by @code{BYTES_BIG_ENDIAN}.
936 @defmac BYTES_BIG_ENDIAN
937 Define this macro to have the value 1 if the most significant byte in a
938 word has the lowest number. This macro need not be a constant.
941 @defmac WORDS_BIG_ENDIAN
942 Define this macro to have the value 1 if, in a multiword object, the
943 most significant word has the lowest number. This applies to both
944 memory locations and registers; GCC fundamentally assumes that the
945 order of words in memory is the same as the order in registers. This
946 macro need not be a constant.
949 @defmac LIBGCC2_WORDS_BIG_ENDIAN
950 Define this macro if @code{WORDS_BIG_ENDIAN} is not constant. This must be a
951 constant value with the same meaning as @code{WORDS_BIG_ENDIAN}, which will be
952 used only when compiling @file{libgcc2.c}. Typically the value will be set
953 based on preprocessor defines.
956 @defmac FLOAT_WORDS_BIG_ENDIAN
957 Define this macro to have the value 1 if @code{DFmode}, @code{XFmode} or
958 @code{TFmode} floating point numbers are stored in memory with the word
959 containing the sign bit at the lowest address; otherwise define it to
960 have the value 0. This macro need not be a constant.
962 You need not define this macro if the ordering is the same as for
966 @defmac BITS_PER_UNIT
967 Define this macro to be the number of bits in an addressable storage
968 unit (byte). If you do not define this macro the default is 8.
971 @defmac BITS_PER_WORD
972 Number of bits in a word. If you do not define this macro, the default
973 is @code{BITS_PER_UNIT * UNITS_PER_WORD}.
976 @defmac MAX_BITS_PER_WORD
977 Maximum number of bits in a word. If this is undefined, the default is
978 @code{BITS_PER_WORD}. Otherwise, it is the constant value that is the
979 largest value that @code{BITS_PER_WORD} can have at run-time.
982 @defmac UNITS_PER_WORD
983 Number of storage units in a word; normally the size of a general-purpose
984 register, a power of two from 1 or 8.
987 @defmac MIN_UNITS_PER_WORD
988 Minimum number of units in a word. If this is undefined, the default is
989 @code{UNITS_PER_WORD}. Otherwise, it is the constant value that is the
990 smallest value that @code{UNITS_PER_WORD} can have at run-time.
993 @defmac UNITS_PER_SIMD_WORD (@var{mode})
994 Number of units in the vectors that the vectorizer can produce for
995 scalar mode @var{mode}. The default is equal to @code{UNITS_PER_WORD},
996 because the vectorizer can do some transformations even in absence of
997 specialized @acronym{SIMD} hardware.
1000 @defmac POINTER_SIZE
1001 Width of a pointer, in bits. You must specify a value no wider than the
1002 width of @code{Pmode}. If it is not equal to the width of @code{Pmode},
1003 you must define @code{POINTERS_EXTEND_UNSIGNED}. If you do not specify
1004 a value the default is @code{BITS_PER_WORD}.
1007 @defmac POINTERS_EXTEND_UNSIGNED
1008 A C expression that determines how pointers should be extended from
1009 @code{ptr_mode} to either @code{Pmode} or @code{word_mode}. It is
1010 greater than zero if pointers should be zero-extended, zero if they
1011 should be sign-extended, and negative if some other sort of conversion
1012 is needed. In the last case, the extension is done by the target's
1013 @code{ptr_extend} instruction.
1015 You need not define this macro if the @code{ptr_mode}, @code{Pmode}
1016 and @code{word_mode} are all the same width.
1019 @defmac PROMOTE_MODE (@var{m}, @var{unsignedp}, @var{type})
1020 A macro to update @var{m} and @var{unsignedp} when an object whose type
1021 is @var{type} and which has the specified mode and signedness is to be
1022 stored in a register. This macro is only called when @var{type} is a
1025 On most RISC machines, which only have operations that operate on a full
1026 register, define this macro to set @var{m} to @code{word_mode} if
1027 @var{m} is an integer mode narrower than @code{BITS_PER_WORD}. In most
1028 cases, only integer modes should be widened because wider-precision
1029 floating-point operations are usually more expensive than their narrower
1032 For most machines, the macro definition does not change @var{unsignedp}.
1033 However, some machines, have instructions that preferentially handle
1034 either signed or unsigned quantities of certain modes. For example, on
1035 the DEC Alpha, 32-bit loads from memory and 32-bit add instructions
1036 sign-extend the result to 64 bits. On such machines, set
1037 @var{unsignedp} according to which kind of extension is more efficient.
1039 Do not define this macro if it would never modify @var{m}.
1042 @deftypefn {Target Hook} enum machine_mode TARGET_PROMOTE_FUNCTION_MODE (tree @var{type}, enum machine_mode @var{mode}, int *@var{punsignedp}, tree @var{funtype}, int @var{for_return})
1043 Like @code{PROMOTE_MODE}, but it is applied to outgoing function arguments or
1044 function return values. The target hook should return the new mode
1045 and possibly change @code{*@var{punsignedp}} if the promotion should
1046 change signedness. This function is called only for scalar @emph{or
1049 @var{for_return} allows to distinguish the promotion of arguments and
1050 return values. If it is @code{1}, a return value is being promoted and
1051 @code{TARGET_FUNCTION_VALUE} must perform the same promotions done here.
1052 If it is @code{2}, the returned mode should be that of the register in
1053 which an incoming parameter is copied, or the outgoing result is computed;
1054 then the hook should return the same mode as @code{promote_mode}, though
1055 the signedness may be different.
1057 The default is to not promote arguments and return values. You can
1058 also define the hook to @code{default_promote_function_mode_always_promote}
1059 if you would like to apply the same rules given by @code{PROMOTE_MODE}.
1062 @defmac PARM_BOUNDARY
1063 Normal alignment required for function parameters on the stack, in
1064 bits. All stack parameters receive at least this much alignment
1065 regardless of data type. On most machines, this is the same as the
1069 @defmac STACK_BOUNDARY
1070 Define this macro to the minimum alignment enforced by hardware for the
1071 stack pointer on this machine. The definition is a C expression for the
1072 desired alignment (measured in bits). This value is used as a default
1073 if @code{PREFERRED_STACK_BOUNDARY} is not defined. On most machines,
1074 this should be the same as @code{PARM_BOUNDARY}.
1077 @defmac PREFERRED_STACK_BOUNDARY
1078 Define this macro if you wish to preserve a certain alignment for the
1079 stack pointer, greater than what the hardware enforces. The definition
1080 is a C expression for the desired alignment (measured in bits). This
1081 macro must evaluate to a value equal to or larger than
1082 @code{STACK_BOUNDARY}.
1085 @defmac INCOMING_STACK_BOUNDARY
1086 Define this macro if the incoming stack boundary may be different
1087 from @code{PREFERRED_STACK_BOUNDARY}. This macro must evaluate
1088 to a value equal to or larger than @code{STACK_BOUNDARY}.
1091 @defmac FUNCTION_BOUNDARY
1092 Alignment required for a function entry point, in bits.
1095 @defmac BIGGEST_ALIGNMENT
1096 Biggest alignment that any data type can require on this machine, in
1097 bits. Note that this is not the biggest alignment that is supported,
1098 just the biggest alignment that, when violated, may cause a fault.
1101 @defmac MALLOC_ABI_ALIGNMENT
1102 Alignment, in bits, a C conformant malloc implementation has to
1103 provide. If not defined, the default value is @code{BITS_PER_WORD}.
1106 @defmac ATTRIBUTE_ALIGNED_VALUE
1107 Alignment used by the @code{__attribute__ ((aligned))} construct. If
1108 not defined, the default value is @code{BIGGEST_ALIGNMENT}.
1111 @defmac MINIMUM_ATOMIC_ALIGNMENT
1112 If defined, the smallest alignment, in bits, that can be given to an
1113 object that can be referenced in one operation, without disturbing any
1114 nearby object. Normally, this is @code{BITS_PER_UNIT}, but may be larger
1115 on machines that don't have byte or half-word store operations.
1118 @defmac BIGGEST_FIELD_ALIGNMENT
1119 Biggest alignment that any structure or union field can require on this
1120 machine, in bits. If defined, this overrides @code{BIGGEST_ALIGNMENT} for
1121 structure and union fields only, unless the field alignment has been set
1122 by the @code{__attribute__ ((aligned (@var{n})))} construct.
1125 @defmac ADJUST_FIELD_ALIGN (@var{field}, @var{computed})
1126 An expression for the alignment of a structure field @var{field} if the
1127 alignment computed in the usual way (including applying of
1128 @code{BIGGEST_ALIGNMENT} and @code{BIGGEST_FIELD_ALIGNMENT} to the
1129 alignment) is @var{computed}. It overrides alignment only if the
1130 field alignment has not been set by the
1131 @code{__attribute__ ((aligned (@var{n})))} construct.
1134 @defmac MAX_STACK_ALIGNMENT
1135 Biggest stack alignment guaranteed by the backend. Use this macro
1136 to specify the maximum alignment of a variable on stack.
1138 If not defined, the default value is @code{STACK_BOUNDARY}.
1140 @c FIXME: The default should be @code{PREFERRED_STACK_BOUNDARY}.
1141 @c But the fix for PR 32893 indicates that we can only guarantee
1142 @c maximum stack alignment on stack up to @code{STACK_BOUNDARY}, not
1143 @c @code{PREFERRED_STACK_BOUNDARY}, if stack alignment isn't supported.
1146 @defmac MAX_OFILE_ALIGNMENT
1147 Biggest alignment supported by the object file format of this machine.
1148 Use this macro to limit the alignment which can be specified using the
1149 @code{__attribute__ ((aligned (@var{n})))} construct. If not defined,
1150 the default value is @code{BIGGEST_ALIGNMENT}.
1152 On systems that use ELF, the default (in @file{config/elfos.h}) is
1153 the largest supported 32-bit ELF section alignment representable on
1154 a 32-bit host e.g. @samp{(((unsigned HOST_WIDEST_INT) 1 << 28) * 8)}.
1155 On 32-bit ELF the largest supported section alignment in bits is
1156 @samp{(0x80000000 * 8)}, but this is not representable on 32-bit hosts.
1159 @defmac DATA_ALIGNMENT (@var{type}, @var{basic-align})
1160 If defined, a C expression to compute the alignment for a variable in
1161 the static store. @var{type} is the data type, and @var{basic-align} is
1162 the alignment that the object would ordinarily have. The value of this
1163 macro is used instead of that alignment to align the object.
1165 If this macro is not defined, then @var{basic-align} is used.
1168 One use of this macro is to increase alignment of medium-size data to
1169 make it all fit in fewer cache lines. Another is to cause character
1170 arrays to be word-aligned so that @code{strcpy} calls that copy
1171 constants to character arrays can be done inline.
1174 @defmac CONSTANT_ALIGNMENT (@var{constant}, @var{basic-align})
1175 If defined, a C expression to compute the alignment given to a constant
1176 that is being placed in memory. @var{constant} is the constant and
1177 @var{basic-align} is the alignment that the object would ordinarily
1178 have. The value of this macro is used instead of that alignment to
1181 If this macro is not defined, then @var{basic-align} is used.
1183 The typical use of this macro is to increase alignment for string
1184 constants to be word aligned so that @code{strcpy} calls that copy
1185 constants can be done inline.
1188 @defmac LOCAL_ALIGNMENT (@var{type}, @var{basic-align})
1189 If defined, a C expression to compute the alignment for a variable in
1190 the local store. @var{type} is the data type, and @var{basic-align} is
1191 the alignment that the object would ordinarily have. The value of this
1192 macro is used instead of that alignment to align the object.
1194 If this macro is not defined, then @var{basic-align} is used.
1196 One use of this macro is to increase alignment of medium-size data to
1197 make it all fit in fewer cache lines.
1200 @defmac STACK_SLOT_ALIGNMENT (@var{type}, @var{mode}, @var{basic-align})
1201 If defined, a C expression to compute the alignment for stack slot.
1202 @var{type} is the data type, @var{mode} is the widest mode available,
1203 and @var{basic-align} is the alignment that the slot would ordinarily
1204 have. The value of this macro is used instead of that alignment to
1207 If this macro is not defined, then @var{basic-align} is used when
1208 @var{type} is @code{NULL}. Otherwise, @code{LOCAL_ALIGNMENT} will
1211 This macro is to set alignment of stack slot to the maximum alignment
1212 of all possible modes which the slot may have.
1215 @defmac LOCAL_DECL_ALIGNMENT (@var{decl})
1216 If defined, a C expression to compute the alignment for a local
1217 variable @var{decl}.
1219 If this macro is not defined, then
1220 @code{LOCAL_ALIGNMENT (TREE_TYPE (@var{decl}), DECL_ALIGN (@var{decl}))}
1223 One use of this macro is to increase alignment of medium-size data to
1224 make it all fit in fewer cache lines.
1227 @defmac MINIMUM_ALIGNMENT (@var{exp}, @var{mode}, @var{align})
1228 If defined, a C expression to compute the minimum required alignment
1229 for dynamic stack realignment purposes for @var{exp} (a type or decl),
1230 @var{mode}, assuming normal alignment @var{align}.
1232 If this macro is not defined, then @var{align} will be used.
1235 @defmac EMPTY_FIELD_BOUNDARY
1236 Alignment in bits to be given to a structure bit-field that follows an
1237 empty field such as @code{int : 0;}.
1239 If @code{PCC_BITFIELD_TYPE_MATTERS} is true, it overrides this macro.
1242 @defmac STRUCTURE_SIZE_BOUNDARY
1243 Number of bits which any structure or union's size must be a multiple of.
1244 Each structure or union's size is rounded up to a multiple of this.
1246 If you do not define this macro, the default is the same as
1247 @code{BITS_PER_UNIT}.
1250 @defmac STRICT_ALIGNMENT
1251 Define this macro to be the value 1 if instructions will fail to work
1252 if given data not on the nominal alignment. If instructions will merely
1253 go slower in that case, define this macro as 0.
1256 @defmac PCC_BITFIELD_TYPE_MATTERS
1257 Define this if you wish to imitate the way many other C compilers handle
1258 alignment of bit-fields and the structures that contain them.
1260 The behavior is that the type written for a named bit-field (@code{int},
1261 @code{short}, or other integer type) imposes an alignment for the entire
1262 structure, as if the structure really did contain an ordinary field of
1263 that type. In addition, the bit-field is placed within the structure so
1264 that it would fit within such a field, not crossing a boundary for it.
1266 Thus, on most machines, a named bit-field whose type is written as
1267 @code{int} would not cross a four-byte boundary, and would force
1268 four-byte alignment for the whole structure. (The alignment used may
1269 not be four bytes; it is controlled by the other alignment parameters.)
1271 An unnamed bit-field will not affect the alignment of the containing
1274 If the macro is defined, its definition should be a C expression;
1275 a nonzero value for the expression enables this behavior.
1277 Note that if this macro is not defined, or its value is zero, some
1278 bit-fields may cross more than one alignment boundary. The compiler can
1279 support such references if there are @samp{insv}, @samp{extv}, and
1280 @samp{extzv} insns that can directly reference memory.
1282 The other known way of making bit-fields work is to define
1283 @code{STRUCTURE_SIZE_BOUNDARY} as large as @code{BIGGEST_ALIGNMENT}.
1284 Then every structure can be accessed with fullwords.
1286 Unless the machine has bit-field instructions or you define
1287 @code{STRUCTURE_SIZE_BOUNDARY} that way, you must define
1288 @code{PCC_BITFIELD_TYPE_MATTERS} to have a nonzero value.
1290 If your aim is to make GCC use the same conventions for laying out
1291 bit-fields as are used by another compiler, here is how to investigate
1292 what the other compiler does. Compile and run this program:
1311 printf ("Size of foo1 is %d\n",
1312 sizeof (struct foo1));
1313 printf ("Size of foo2 is %d\n",
1314 sizeof (struct foo2));
1319 If this prints 2 and 5, then the compiler's behavior is what you would
1320 get from @code{PCC_BITFIELD_TYPE_MATTERS}.
1323 @defmac BITFIELD_NBYTES_LIMITED
1324 Like @code{PCC_BITFIELD_TYPE_MATTERS} except that its effect is limited
1325 to aligning a bit-field within the structure.
1328 @deftypefn {Target Hook} bool TARGET_ALIGN_ANON_BITFIELD (void)
1329 When @code{PCC_BITFIELD_TYPE_MATTERS} is true this hook will determine
1330 whether unnamed bitfields affect the alignment of the containing
1331 structure. The hook should return true if the structure should inherit
1332 the alignment requirements of an unnamed bitfield's type.
1335 @deftypefn {Target Hook} bool TARGET_NARROW_VOLATILE_BITFIELD (void)
1336 This target hook should return @code{true} if accesses to volatile bitfields
1337 should use the narrowest mode possible. It should return @code{false} if
1338 these accesses should use the bitfield container type.
1340 The default is @code{!TARGET_STRICT_ALIGN}.
1343 @defmac MEMBER_TYPE_FORCES_BLK (@var{field}, @var{mode})
1344 Return 1 if a structure or array containing @var{field} should be accessed using
1347 If @var{field} is the only field in the structure, @var{mode} is its
1348 mode, otherwise @var{mode} is VOIDmode. @var{mode} is provided in the
1349 case where structures of one field would require the structure's mode to
1350 retain the field's mode.
1352 Normally, this is not needed.
1355 @defmac ROUND_TYPE_ALIGN (@var{type}, @var{computed}, @var{specified})
1356 Define this macro as an expression for the alignment of a type (given
1357 by @var{type} as a tree node) if the alignment computed in the usual
1358 way is @var{computed} and the alignment explicitly specified was
1361 The default is to use @var{specified} if it is larger; otherwise, use
1362 the smaller of @var{computed} and @code{BIGGEST_ALIGNMENT}
1365 @defmac MAX_FIXED_MODE_SIZE
1366 An integer expression for the size in bits of the largest integer
1367 machine mode that should actually be used. All integer machine modes of
1368 this size or smaller can be used for structures and unions with the
1369 appropriate sizes. If this macro is undefined, @code{GET_MODE_BITSIZE
1370 (DImode)} is assumed.
1373 @defmac STACK_SAVEAREA_MODE (@var{save_level})
1374 If defined, an expression of type @code{enum machine_mode} that
1375 specifies the mode of the save area operand of a
1376 @code{save_stack_@var{level}} named pattern (@pxref{Standard Names}).
1377 @var{save_level} is one of @code{SAVE_BLOCK}, @code{SAVE_FUNCTION}, or
1378 @code{SAVE_NONLOCAL} and selects which of the three named patterns is
1379 having its mode specified.
1381 You need not define this macro if it always returns @code{Pmode}. You
1382 would most commonly define this macro if the
1383 @code{save_stack_@var{level}} patterns need to support both a 32- and a
1387 @defmac STACK_SIZE_MODE
1388 If defined, an expression of type @code{enum machine_mode} that
1389 specifies the mode of the size increment operand of an
1390 @code{allocate_stack} named pattern (@pxref{Standard Names}).
1392 You need not define this macro if it always returns @code{word_mode}.
1393 You would most commonly define this macro if the @code{allocate_stack}
1394 pattern needs to support both a 32- and a 64-bit mode.
1397 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_CMP_RETURN_MODE ()
1398 This target hook should return the mode to be used for the return value
1399 of compare instructions expanded to libgcc calls. If not defined
1400 @code{word_mode} is returned which is the right choice for a majority of
1404 @deftypefn {Target Hook} {enum machine_mode} TARGET_LIBGCC_SHIFT_COUNT_MODE ()
1405 This target hook should return the mode to be used for the shift count operand
1406 of shift instructions expanded to libgcc calls. If not defined
1407 @code{word_mode} is returned which is the right choice for a majority of
1411 @defmac ROUND_TOWARDS_ZERO
1412 If defined, this macro should be true if the prevailing rounding
1413 mode is towards zero.
1415 Defining this macro only affects the way @file{libgcc.a} emulates
1416 floating-point arithmetic.
1418 Not defining this macro is equivalent to returning zero.
1421 @defmac LARGEST_EXPONENT_IS_NORMAL (@var{size})
1422 This macro should return true if floats with @var{size}
1423 bits do not have a NaN or infinity representation, but use the largest
1424 exponent for normal numbers instead.
1426 Defining this macro only affects the way @file{libgcc.a} emulates
1427 floating-point arithmetic.
1429 The default definition of this macro returns false for all sizes.
1432 @deftypefn {Target Hook} bool TARGET_MS_BITFIELD_LAYOUT_P (tree @var{record_type})
1433 This target hook returns @code{true} if bit-fields in the given
1434 @var{record_type} are to be laid out following the rules of Microsoft
1435 Visual C/C++, namely: (i) a bit-field won't share the same storage
1436 unit with the previous bit-field if their underlying types have
1437 different sizes, and the bit-field will be aligned to the highest
1438 alignment of the underlying types of itself and of the previous
1439 bit-field; (ii) a zero-sized bit-field will affect the alignment of
1440 the whole enclosing structure, even if it is unnamed; except that
1441 (iii) a zero-sized bit-field will be disregarded unless it follows
1442 another bit-field of nonzero size. If this hook returns @code{true},
1443 other macros that control bit-field layout are ignored.
1445 When a bit-field is inserted into a packed record, the whole size
1446 of the underlying type is used by one or more same-size adjacent
1447 bit-fields (that is, if its long:3, 32 bits is used in the record,
1448 and any additional adjacent long bit-fields are packed into the same
1449 chunk of 32 bits. However, if the size changes, a new field of that
1450 size is allocated). In an unpacked record, this is the same as using
1451 alignment, but not equivalent when packing.
1453 If both MS bit-fields and @samp{__attribute__((packed))} are used,
1454 the latter will take precedence. If @samp{__attribute__((packed))} is
1455 used on a single field when MS bit-fields are in use, it will take
1456 precedence for that field, but the alignment of the rest of the structure
1457 may affect its placement.
1460 @deftypefn {Target Hook} {bool} TARGET_DECIMAL_FLOAT_SUPPORTED_P (void)
1461 Returns true if the target supports decimal floating point.
1464 @deftypefn {Target Hook} {bool} TARGET_FIXED_POINT_SUPPORTED_P (void)
1465 Returns true if the target supports fixed-point arithmetic.
1468 @deftypefn {Target Hook} void TARGET_EXPAND_TO_RTL_HOOK (void)
1469 This hook is called just before expansion into rtl, allowing the target
1470 to perform additional initializations or analysis before the expansion.
1471 For example, the rs6000 port uses it to allocate a scratch stack slot
1472 for use in copying SDmode values between memory and floating point
1473 registers whenever the function being expanded has any SDmode
1477 @deftypefn {Target Hook} void TARGET_INSTANTIATE_DECLS (void)
1478 This hook allows the backend to perform additional instantiations on rtl
1479 that are not actually in any insns yet, but will be later.
1482 @deftypefn {Target Hook} {const char *} TARGET_MANGLE_TYPE (tree @var{type})
1483 If your target defines any fundamental types, or any types your target
1484 uses should be mangled differently from the default, define this hook
1485 to return the appropriate encoding for these types as part of a C++
1486 mangled name. The @var{type} argument is the tree structure representing
1487 the type to be mangled. The hook may be applied to trees which are
1488 not target-specific fundamental types; it should return @code{NULL}
1489 for all such types, as well as arguments it does not recognize. If the
1490 return value is not @code{NULL}, it must point to a statically-allocated
1493 Target-specific fundamental types might be new fundamental types or
1494 qualified versions of ordinary fundamental types. Encode new
1495 fundamental types as @samp{@w{u @var{n} @var{name}}}, where @var{name}
1496 is the name used for the type in source code, and @var{n} is the
1497 length of @var{name} in decimal. Encode qualified versions of
1498 ordinary types as @samp{@w{U @var{n} @var{name} @var{code}}}, where
1499 @var{name} is the name used for the type qualifier in source code,
1500 @var{n} is the length of @var{name} as above, and @var{code} is the
1501 code used to represent the unqualified version of this type. (See
1502 @code{write_builtin_type} in @file{cp/mangle.c} for the list of
1503 codes.) In both cases the spaces are for clarity; do not include any
1504 spaces in your string.
1506 This hook is applied to types prior to typedef resolution. If the mangled
1507 name for a particular type depends only on that type's main variant, you
1508 can perform typedef resolution yourself using @code{TYPE_MAIN_VARIANT}
1511 The default version of this hook always returns @code{NULL}, which is
1512 appropriate for a target that does not define any new fundamental
1517 @section Layout of Source Language Data Types
1519 These macros define the sizes and other characteristics of the standard
1520 basic data types used in programs being compiled. Unlike the macros in
1521 the previous section, these apply to specific features of C and related
1522 languages, rather than to fundamental aspects of storage layout.
1524 @defmac INT_TYPE_SIZE
1525 A C expression for the size in bits of the type @code{int} on the
1526 target machine. If you don't define this, the default is one word.
1529 @defmac SHORT_TYPE_SIZE
1530 A C expression for the size in bits of the type @code{short} on the
1531 target machine. If you don't define this, the default is half a word.
1532 (If this would be less than one storage unit, it is rounded up to one
1536 @defmac LONG_TYPE_SIZE
1537 A C expression for the size in bits of the type @code{long} on the
1538 target machine. If you don't define this, the default is one word.
1541 @defmac ADA_LONG_TYPE_SIZE
1542 On some machines, the size used for the Ada equivalent of the type
1543 @code{long} by a native Ada compiler differs from that used by C@. In
1544 that situation, define this macro to be a C expression to be used for
1545 the size of that type. If you don't define this, the default is the
1546 value of @code{LONG_TYPE_SIZE}.
1549 @defmac LONG_LONG_TYPE_SIZE
1550 A C expression for the size in bits of the type @code{long long} on the
1551 target machine. If you don't define this, the default is two
1552 words. If you want to support GNU Ada on your machine, the value of this
1553 macro must be at least 64.
1556 @defmac CHAR_TYPE_SIZE
1557 A C expression for the size in bits of the type @code{char} on the
1558 target machine. If you don't define this, the default is
1559 @code{BITS_PER_UNIT}.
1562 @defmac BOOL_TYPE_SIZE
1563 A C expression for the size in bits of the C++ type @code{bool} and
1564 C99 type @code{_Bool} on the target machine. If you don't define
1565 this, and you probably shouldn't, the default is @code{CHAR_TYPE_SIZE}.
1568 @defmac FLOAT_TYPE_SIZE
1569 A C expression for the size in bits of the type @code{float} on the
1570 target machine. If you don't define this, the default is one word.
1573 @defmac DOUBLE_TYPE_SIZE
1574 A C expression for the size in bits of the type @code{double} on the
1575 target machine. If you don't define this, the default is two
1579 @defmac LONG_DOUBLE_TYPE_SIZE
1580 A C expression for the size in bits of the type @code{long double} on
1581 the target machine. If you don't define this, the default is two
1585 @defmac SHORT_FRACT_TYPE_SIZE
1586 A C expression for the size in bits of the type @code{short _Fract} on
1587 the target machine. If you don't define this, the default is
1588 @code{BITS_PER_UNIT}.
1591 @defmac FRACT_TYPE_SIZE
1592 A C expression for the size in bits of the type @code{_Fract} on
1593 the target machine. If you don't define this, the default is
1594 @code{BITS_PER_UNIT * 2}.
1597 @defmac LONG_FRACT_TYPE_SIZE
1598 A C expression for the size in bits of the type @code{long _Fract} on
1599 the target machine. If you don't define this, the default is
1600 @code{BITS_PER_UNIT * 4}.
1603 @defmac LONG_LONG_FRACT_TYPE_SIZE
1604 A C expression for the size in bits of the type @code{long long _Fract} on
1605 the target machine. If you don't define this, the default is
1606 @code{BITS_PER_UNIT * 8}.
1609 @defmac SHORT_ACCUM_TYPE_SIZE
1610 A C expression for the size in bits of the type @code{short _Accum} on
1611 the target machine. If you don't define this, the default is
1612 @code{BITS_PER_UNIT * 2}.
1615 @defmac ACCUM_TYPE_SIZE
1616 A C expression for the size in bits of the type @code{_Accum} on
1617 the target machine. If you don't define this, the default is
1618 @code{BITS_PER_UNIT * 4}.
1621 @defmac LONG_ACCUM_TYPE_SIZE
1622 A C expression for the size in bits of the type @code{long _Accum} on
1623 the target machine. If you don't define this, the default is
1624 @code{BITS_PER_UNIT * 8}.
1627 @defmac LONG_LONG_ACCUM_TYPE_SIZE
1628 A C expression for the size in bits of the type @code{long long _Accum} on
1629 the target machine. If you don't define this, the default is
1630 @code{BITS_PER_UNIT * 16}.
1633 @defmac LIBGCC2_LONG_DOUBLE_TYPE_SIZE
1634 Define this macro if @code{LONG_DOUBLE_TYPE_SIZE} is not constant or
1635 if you want routines in @file{libgcc2.a} for a size other than
1636 @code{LONG_DOUBLE_TYPE_SIZE}. If you don't define this, the
1637 default is @code{LONG_DOUBLE_TYPE_SIZE}.
1640 @defmac LIBGCC2_HAS_DF_MODE
1641 Define this macro if neither @code{LIBGCC2_DOUBLE_TYPE_SIZE} nor
1642 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is
1643 @code{DFmode} but you want @code{DFmode} routines in @file{libgcc2.a}
1644 anyway. If you don't define this and either @code{LIBGCC2_DOUBLE_TYPE_SIZE}
1645 or @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64 then the default is 1,
1649 @defmac LIBGCC2_HAS_XF_MODE
1650 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1651 @code{XFmode} but you want @code{XFmode} routines in @file{libgcc2.a}
1652 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1653 is 80 then the default is 1, otherwise it is 0.
1656 @defmac LIBGCC2_HAS_TF_MODE
1657 Define this macro if @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is not
1658 @code{TFmode} but you want @code{TFmode} routines in @file{libgcc2.a}
1659 anyway. If you don't define this and @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE}
1660 is 128 then the default is 1, otherwise it is 0.
1667 Define these macros to be the size in bits of the mantissa of
1668 @code{SFmode}, @code{DFmode}, @code{XFmode} and @code{TFmode} values,
1669 if the defaults in @file{libgcc2.h} are inappropriate. By default,
1670 @code{FLT_MANT_DIG} is used for @code{SF_SIZE}, @code{LDBL_MANT_DIG}
1671 for @code{XF_SIZE} and @code{TF_SIZE}, and @code{DBL_MANT_DIG} or
1672 @code{LDBL_MANT_DIG} for @code{DF_SIZE} according to whether
1673 @code{LIBGCC2_DOUBLE_TYPE_SIZE} or
1674 @code{LIBGCC2_LONG_DOUBLE_TYPE_SIZE} is 64.
1677 @defmac TARGET_FLT_EVAL_METHOD
1678 A C expression for the value for @code{FLT_EVAL_METHOD} in @file{float.h},
1679 assuming, if applicable, that the floating-point control word is in its
1680 default state. If you do not define this macro the value of
1681 @code{FLT_EVAL_METHOD} will be zero.
1684 @defmac WIDEST_HARDWARE_FP_SIZE
1685 A C expression for the size in bits of the widest floating-point format
1686 supported by the hardware. If you define this macro, you must specify a
1687 value less than or equal to the value of @code{LONG_DOUBLE_TYPE_SIZE}.
1688 If you do not define this macro, the value of @code{LONG_DOUBLE_TYPE_SIZE}
1692 @defmac DEFAULT_SIGNED_CHAR
1693 An expression whose value is 1 or 0, according to whether the type
1694 @code{char} should be signed or unsigned by default. The user can
1695 always override this default with the options @option{-fsigned-char}
1696 and @option{-funsigned-char}.
1699 @deftypefn {Target Hook} bool TARGET_DEFAULT_SHORT_ENUMS (void)
1700 This target hook should return true if the compiler should give an
1701 @code{enum} type only as many bytes as it takes to represent the range
1702 of possible values of that type. It should return false if all
1703 @code{enum} types should be allocated like @code{int}.
1705 The default is to return false.
1709 A C expression for a string describing the name of the data type to use
1710 for size values. The typedef name @code{size_t} is defined using the
1711 contents of the string.
1713 The string can contain more than one keyword. If so, separate them with
1714 spaces, and write first any length keyword, then @code{unsigned} if
1715 appropriate, and finally @code{int}. The string must exactly match one
1716 of the data type names defined in the function
1717 @code{init_decl_processing} in the file @file{c-decl.c}. You may not
1718 omit @code{int} or change the order---that would cause the compiler to
1721 If you don't define this macro, the default is @code{"long unsigned
1725 @defmac PTRDIFF_TYPE
1726 A C expression for a string describing the name of the data type to use
1727 for the result of subtracting two pointers. The typedef name
1728 @code{ptrdiff_t} is defined using the contents of the string. See
1729 @code{SIZE_TYPE} above for more information.
1731 If you don't define this macro, the default is @code{"long int"}.
1735 A C expression for a string describing the name of the data type to use
1736 for wide characters. The typedef name @code{wchar_t} is defined using
1737 the contents of the string. See @code{SIZE_TYPE} above for more
1740 If you don't define this macro, the default is @code{"int"}.
1743 @defmac WCHAR_TYPE_SIZE
1744 A C expression for the size in bits of the data type for wide
1745 characters. This is used in @code{cpp}, which cannot make use of
1750 A C expression for a string describing the name of the data type to
1751 use for wide characters passed to @code{printf} and returned from
1752 @code{getwc}. The typedef name @code{wint_t} is defined using the
1753 contents of the string. See @code{SIZE_TYPE} above for more
1756 If you don't define this macro, the default is @code{"unsigned int"}.
1760 A C expression for a string describing the name of the data type that
1761 can represent any value of any standard or extended signed integer type.
1762 The typedef name @code{intmax_t} is defined using the contents of the
1763 string. See @code{SIZE_TYPE} above for more information.
1765 If you don't define this macro, the default is the first of
1766 @code{"int"}, @code{"long int"}, or @code{"long long int"} that has as
1767 much precision as @code{long long int}.
1770 @defmac UINTMAX_TYPE
1771 A C expression for a string describing the name of the data type that
1772 can represent any value of any standard or extended unsigned integer
1773 type. The typedef name @code{uintmax_t} is defined using the contents
1774 of the string. See @code{SIZE_TYPE} above for more information.
1776 If you don't define this macro, the default is the first of
1777 @code{"unsigned int"}, @code{"long unsigned int"}, or @code{"long long
1778 unsigned int"} that has as much precision as @code{long long unsigned
1782 @defmac SIG_ATOMIC_TYPE
1788 @defmacx UINT16_TYPE
1789 @defmacx UINT32_TYPE
1790 @defmacx UINT64_TYPE
1791 @defmacx INT_LEAST8_TYPE
1792 @defmacx INT_LEAST16_TYPE
1793 @defmacx INT_LEAST32_TYPE
1794 @defmacx INT_LEAST64_TYPE
1795 @defmacx UINT_LEAST8_TYPE
1796 @defmacx UINT_LEAST16_TYPE
1797 @defmacx UINT_LEAST32_TYPE
1798 @defmacx UINT_LEAST64_TYPE
1799 @defmacx INT_FAST8_TYPE
1800 @defmacx INT_FAST16_TYPE
1801 @defmacx INT_FAST32_TYPE
1802 @defmacx INT_FAST64_TYPE
1803 @defmacx UINT_FAST8_TYPE
1804 @defmacx UINT_FAST16_TYPE
1805 @defmacx UINT_FAST32_TYPE
1806 @defmacx UINT_FAST64_TYPE
1807 @defmacx INTPTR_TYPE
1808 @defmacx UINTPTR_TYPE
1809 C expressions for the standard types @code{sig_atomic_t},
1810 @code{int8_t}, @code{int16_t}, @code{int32_t}, @code{int64_t},
1811 @code{uint8_t}, @code{uint16_t}, @code{uint32_t}, @code{uint64_t},
1812 @code{int_least8_t}, @code{int_least16_t}, @code{int_least32_t},
1813 @code{int_least64_t}, @code{uint_least8_t}, @code{uint_least16_t},
1814 @code{uint_least32_t}, @code{uint_least64_t}, @code{int_fast8_t},
1815 @code{int_fast16_t}, @code{int_fast32_t}, @code{int_fast64_t},
1816 @code{uint_fast8_t}, @code{uint_fast16_t}, @code{uint_fast32_t},
1817 @code{uint_fast64_t}, @code{intptr_t}, and @code{uintptr_t}. See
1818 @code{SIZE_TYPE} above for more information.
1820 If any of these macros evaluates to a null pointer, the corresponding
1821 type is not supported; if GCC is configured to provide
1822 @code{<stdint.h>} in such a case, the header provided may not conform
1823 to C99, depending on the type in question. The defaults for all of
1824 these macros are null pointers.
1827 @defmac TARGET_PTRMEMFUNC_VBIT_LOCATION
1828 The C++ compiler represents a pointer-to-member-function with a struct
1835 ptrdiff_t vtable_index;
1842 The C++ compiler must use one bit to indicate whether the function that
1843 will be called through a pointer-to-member-function is virtual.
1844 Normally, we assume that the low-order bit of a function pointer must
1845 always be zero. Then, by ensuring that the vtable_index is odd, we can
1846 distinguish which variant of the union is in use. But, on some
1847 platforms function pointers can be odd, and so this doesn't work. In
1848 that case, we use the low-order bit of the @code{delta} field, and shift
1849 the remainder of the @code{delta} field to the left.
1851 GCC will automatically make the right selection about where to store
1852 this bit using the @code{FUNCTION_BOUNDARY} setting for your platform.
1853 However, some platforms such as ARM/Thumb have @code{FUNCTION_BOUNDARY}
1854 set such that functions always start at even addresses, but the lowest
1855 bit of pointers to functions indicate whether the function at that
1856 address is in ARM or Thumb mode. If this is the case of your
1857 architecture, you should define this macro to
1858 @code{ptrmemfunc_vbit_in_delta}.
1860 In general, you should not have to define this macro. On architectures
1861 in which function addresses are always even, according to
1862 @code{FUNCTION_BOUNDARY}, GCC will automatically define this macro to
1863 @code{ptrmemfunc_vbit_in_pfn}.
1866 @defmac TARGET_VTABLE_USES_DESCRIPTORS
1867 Normally, the C++ compiler uses function pointers in vtables. This
1868 macro allows the target to change to use ``function descriptors''
1869 instead. Function descriptors are found on targets for whom a
1870 function pointer is actually a small data structure. Normally the
1871 data structure consists of the actual code address plus a data
1872 pointer to which the function's data is relative.
1874 If vtables are used, the value of this macro should be the number
1875 of words that the function descriptor occupies.
1878 @defmac TARGET_VTABLE_ENTRY_ALIGN
1879 By default, the vtable entries are void pointers, the so the alignment
1880 is the same as pointer alignment. The value of this macro specifies
1881 the alignment of the vtable entry in bits. It should be defined only
1882 when special alignment is necessary. */
1885 @defmac TARGET_VTABLE_DATA_ENTRY_DISTANCE
1886 There are a few non-descriptor entries in the vtable at offsets below
1887 zero. If these entries must be padded (say, to preserve the alignment
1888 specified by @code{TARGET_VTABLE_ENTRY_ALIGN}), set this to the number
1889 of words in each data entry.
1893 @section Register Usage
1894 @cindex register usage
1896 This section explains how to describe what registers the target machine
1897 has, and how (in general) they can be used.
1899 The description of which registers a specific instruction can use is
1900 done with register classes; see @ref{Register Classes}. For information
1901 on using registers to access a stack frame, see @ref{Frame Registers}.
1902 For passing values in registers, see @ref{Register Arguments}.
1903 For returning values in registers, see @ref{Scalar Return}.
1906 * Register Basics:: Number and kinds of registers.
1907 * Allocation Order:: Order in which registers are allocated.
1908 * Values in Registers:: What kinds of values each reg can hold.
1909 * Leaf Functions:: Renumbering registers for leaf functions.
1910 * Stack Registers:: Handling a register stack such as 80387.
1913 @node Register Basics
1914 @subsection Basic Characteristics of Registers
1916 @c prevent bad page break with this line
1917 Registers have various characteristics.
1919 @defmac FIRST_PSEUDO_REGISTER
1920 Number of hardware registers known to the compiler. They receive
1921 numbers 0 through @code{FIRST_PSEUDO_REGISTER-1}; thus, the first
1922 pseudo register's number really is assigned the number
1923 @code{FIRST_PSEUDO_REGISTER}.
1926 @defmac FIXED_REGISTERS
1927 @cindex fixed register
1928 An initializer that says which registers are used for fixed purposes
1929 all throughout the compiled code and are therefore not available for
1930 general allocation. These would include the stack pointer, the frame
1931 pointer (except on machines where that can be used as a general
1932 register when no frame pointer is needed), the program counter on
1933 machines where that is considered one of the addressable registers,
1934 and any other numbered register with a standard use.
1936 This information is expressed as a sequence of numbers, separated by
1937 commas and surrounded by braces. The @var{n}th number is 1 if
1938 register @var{n} is fixed, 0 otherwise.
1940 The table initialized from this macro, and the table initialized by
1941 the following one, may be overridden at run time either automatically,
1942 by the actions of the macro @code{CONDITIONAL_REGISTER_USAGE}, or by
1943 the user with the command options @option{-ffixed-@var{reg}},
1944 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}.
1947 @defmac CALL_USED_REGISTERS
1948 @cindex call-used register
1949 @cindex call-clobbered register
1950 @cindex call-saved register
1951 Like @code{FIXED_REGISTERS} but has 1 for each register that is
1952 clobbered (in general) by function calls as well as for fixed
1953 registers. This macro therefore identifies the registers that are not
1954 available for general allocation of values that must live across
1957 If a register has 0 in @code{CALL_USED_REGISTERS}, the compiler
1958 automatically saves it on function entry and restores it on function
1959 exit, if the register is used within the function.
1962 @defmac CALL_REALLY_USED_REGISTERS
1963 @cindex call-used register
1964 @cindex call-clobbered register
1965 @cindex call-saved register
1966 Like @code{CALL_USED_REGISTERS} except this macro doesn't require
1967 that the entire set of @code{FIXED_REGISTERS} be included.
1968 (@code{CALL_USED_REGISTERS} must be a superset of @code{FIXED_REGISTERS}).
1969 This macro is optional. If not specified, it defaults to the value
1970 of @code{CALL_USED_REGISTERS}.
1973 @defmac HARD_REGNO_CALL_PART_CLOBBERED (@var{regno}, @var{mode})
1974 @cindex call-used register
1975 @cindex call-clobbered register
1976 @cindex call-saved register
1977 A C expression that is nonzero if it is not permissible to store a
1978 value of mode @var{mode} in hard register number @var{regno} across a
1979 call without some part of it being clobbered. For most machines this
1980 macro need not be defined. It is only required for machines that do not
1981 preserve the entire contents of a register across a call.
1985 @findex call_used_regs
1988 @findex reg_class_contents
1989 @defmac CONDITIONAL_REGISTER_USAGE
1990 Zero or more C statements that may conditionally modify five variables
1991 @code{fixed_regs}, @code{call_used_regs}, @code{global_regs},
1992 @code{reg_names}, and @code{reg_class_contents}, to take into account
1993 any dependence of these register sets on target flags. The first three
1994 of these are of type @code{char []} (interpreted as Boolean vectors).
1995 @code{global_regs} is a @code{const char *[]}, and
1996 @code{reg_class_contents} is a @code{HARD_REG_SET}. Before the macro is
1997 called, @code{fixed_regs}, @code{call_used_regs},
1998 @code{reg_class_contents}, and @code{reg_names} have been initialized
1999 from @code{FIXED_REGISTERS}, @code{CALL_USED_REGISTERS},
2000 @code{REG_CLASS_CONTENTS}, and @code{REGISTER_NAMES}, respectively.
2001 @code{global_regs} has been cleared, and any @option{-ffixed-@var{reg}},
2002 @option{-fcall-used-@var{reg}} and @option{-fcall-saved-@var{reg}}
2003 command options have been applied.
2005 You need not define this macro if it has no work to do.
2007 @cindex disabling certain registers
2008 @cindex controlling register usage
2009 If the usage of an entire class of registers depends on the target
2010 flags, you may indicate this to GCC by using this macro to modify
2011 @code{fixed_regs} and @code{call_used_regs} to 1 for each of the
2012 registers in the classes which should not be used by GCC@. Also define
2013 the macro @code{REG_CLASS_FROM_LETTER} / @code{REG_CLASS_FROM_CONSTRAINT}
2014 to return @code{NO_REGS} if it
2015 is called with a letter for a class that shouldn't be used.
2017 (However, if this class is not included in @code{GENERAL_REGS} and all
2018 of the insn patterns whose constraints permit this class are
2019 controlled by target switches, then GCC will automatically avoid using
2020 these registers when the target switches are opposed to them.)
2023 @defmac INCOMING_REGNO (@var{out})
2024 Define this macro if the target machine has register windows. This C
2025 expression returns the register number as seen by the called function
2026 corresponding to the register number @var{out} as seen by the calling
2027 function. Return @var{out} if register number @var{out} is not an
2031 @defmac OUTGOING_REGNO (@var{in})
2032 Define this macro if the target machine has register windows. This C
2033 expression returns the register number as seen by the calling function
2034 corresponding to the register number @var{in} as seen by the called
2035 function. Return @var{in} if register number @var{in} is not an inbound
2039 @defmac LOCAL_REGNO (@var{regno})
2040 Define this macro if the target machine has register windows. This C
2041 expression returns true if the register is call-saved but is in the
2042 register window. Unlike most call-saved registers, such registers
2043 need not be explicitly restored on function exit or during non-local
2048 If the program counter has a register number, define this as that
2049 register number. Otherwise, do not define it.
2052 @node Allocation Order
2053 @subsection Order of Allocation of Registers
2054 @cindex order of register allocation
2055 @cindex register allocation order
2057 @c prevent bad page break with this line
2058 Registers are allocated in order.
2060 @defmac REG_ALLOC_ORDER
2061 If defined, an initializer for a vector of integers, containing the
2062 numbers of hard registers in the order in which GCC should prefer
2063 to use them (from most preferred to least).
2065 If this macro is not defined, registers are used lowest numbered first
2066 (all else being equal).
2068 One use of this macro is on machines where the highest numbered
2069 registers must always be saved and the save-multiple-registers
2070 instruction supports only sequences of consecutive registers. On such
2071 machines, define @code{REG_ALLOC_ORDER} to be an initializer that lists
2072 the highest numbered allocable register first.
2075 @defmac ORDER_REGS_FOR_LOCAL_ALLOC
2076 A C statement (sans semicolon) to choose the order in which to allocate
2077 hard registers for pseudo-registers local to a basic block.
2079 Store the desired register order in the array @code{reg_alloc_order}.
2080 Element 0 should be the register to allocate first; element 1, the next
2081 register; and so on.
2083 The macro body should not assume anything about the contents of
2084 @code{reg_alloc_order} before execution of the macro.
2086 On most machines, it is not necessary to define this macro.
2089 @defmac IRA_HARD_REGNO_ADD_COST_MULTIPLIER (@var{regno})
2090 In some case register allocation order is not enough for the
2091 Integrated Register Allocator (@acronym{IRA}) to generate a good code.
2092 If this macro is defined, it should return a floating point value
2093 based on @var{regno}. The cost of using @var{regno} for a pseudo will
2094 be increased by approximately the pseudo's usage frequency times the
2095 value returned by this macro. Not defining this macro is equivalent
2096 to having it always return @code{0.0}.
2098 On most machines, it is not necessary to define this macro.
2101 @node Values in Registers
2102 @subsection How Values Fit in Registers
2104 This section discusses the macros that describe which kinds of values
2105 (specifically, which machine modes) each register can hold, and how many
2106 consecutive registers are needed for a given mode.
2108 @defmac HARD_REGNO_NREGS (@var{regno}, @var{mode})
2109 A C expression for the number of consecutive hard registers, starting
2110 at register number @var{regno}, required to hold a value of mode
2111 @var{mode}. This macro must never return zero, even if a register
2112 cannot hold the requested mode - indicate that with HARD_REGNO_MODE_OK
2113 and/or CANNOT_CHANGE_MODE_CLASS instead.
2115 On a machine where all registers are exactly one word, a suitable
2116 definition of this macro is
2119 #define HARD_REGNO_NREGS(REGNO, MODE) \
2120 ((GET_MODE_SIZE (MODE) + UNITS_PER_WORD - 1) \
2125 @defmac HARD_REGNO_NREGS_HAS_PADDING (@var{regno}, @var{mode})
2126 A C expression that is nonzero if a value of mode @var{mode}, stored
2127 in memory, ends with padding that causes it to take up more space than
2128 in registers starting at register number @var{regno} (as determined by
2129 multiplying GCC's notion of the size of the register when containing
2130 this mode by the number of registers returned by
2131 @code{HARD_REGNO_NREGS}). By default this is zero.
2133 For example, if a floating-point value is stored in three 32-bit
2134 registers but takes up 128 bits in memory, then this would be
2137 This macros only needs to be defined if there are cases where
2138 @code{subreg_get_info}
2139 would otherwise wrongly determine that a @code{subreg} can be
2140 represented by an offset to the register number, when in fact such a
2141 @code{subreg} would contain some of the padding not stored in
2142 registers and so not be representable.
2145 @defmac HARD_REGNO_NREGS_WITH_PADDING (@var{regno}, @var{mode})
2146 For values of @var{regno} and @var{mode} for which
2147 @code{HARD_REGNO_NREGS_HAS_PADDING} returns nonzero, a C expression
2148 returning the greater number of registers required to hold the value
2149 including any padding. In the example above, the value would be four.
2152 @defmac REGMODE_NATURAL_SIZE (@var{mode})
2153 Define this macro if the natural size of registers that hold values
2154 of mode @var{mode} is not the word size. It is a C expression that
2155 should give the natural size in bytes for the specified mode. It is
2156 used by the register allocator to try to optimize its results. This
2157 happens for example on SPARC 64-bit where the natural size of
2158 floating-point registers is still 32-bit.
2161 @defmac HARD_REGNO_MODE_OK (@var{regno}, @var{mode})
2162 A C expression that is nonzero if it is permissible to store a value
2163 of mode @var{mode} in hard register number @var{regno} (or in several
2164 registers starting with that one). For a machine where all registers
2165 are equivalent, a suitable definition is
2168 #define HARD_REGNO_MODE_OK(REGNO, MODE) 1
2171 You need not include code to check for the numbers of fixed registers,
2172 because the allocation mechanism considers them to be always occupied.
2174 @cindex register pairs
2175 On some machines, double-precision values must be kept in even/odd
2176 register pairs. You can implement that by defining this macro to reject
2177 odd register numbers for such modes.
2179 The minimum requirement for a mode to be OK in a register is that the
2180 @samp{mov@var{mode}} instruction pattern support moves between the
2181 register and other hard register in the same class and that moving a
2182 value into the register and back out not alter it.
2184 Since the same instruction used to move @code{word_mode} will work for
2185 all narrower integer modes, it is not necessary on any machine for
2186 @code{HARD_REGNO_MODE_OK} to distinguish between these modes, provided
2187 you define patterns @samp{movhi}, etc., to take advantage of this. This
2188 is useful because of the interaction between @code{HARD_REGNO_MODE_OK}
2189 and @code{MODES_TIEABLE_P}; it is very desirable for all integer modes
2192 Many machines have special registers for floating point arithmetic.
2193 Often people assume that floating point machine modes are allowed only
2194 in floating point registers. This is not true. Any registers that
2195 can hold integers can safely @emph{hold} a floating point machine
2196 mode, whether or not floating arithmetic can be done on it in those
2197 registers. Integer move instructions can be used to move the values.
2199 On some machines, though, the converse is true: fixed-point machine
2200 modes may not go in floating registers. This is true if the floating
2201 registers normalize any value stored in them, because storing a
2202 non-floating value there would garble it. In this case,
2203 @code{HARD_REGNO_MODE_OK} should reject fixed-point machine modes in
2204 floating registers. But if the floating registers do not automatically
2205 normalize, if you can store any bit pattern in one and retrieve it
2206 unchanged without a trap, then any machine mode may go in a floating
2207 register, so you can define this macro to say so.
2209 The primary significance of special floating registers is rather that
2210 they are the registers acceptable in floating point arithmetic
2211 instructions. However, this is of no concern to
2212 @code{HARD_REGNO_MODE_OK}. You handle it by writing the proper
2213 constraints for those instructions.
2215 On some machines, the floating registers are especially slow to access,
2216 so that it is better to store a value in a stack frame than in such a
2217 register if floating point arithmetic is not being done. As long as the
2218 floating registers are not in class @code{GENERAL_REGS}, they will not
2219 be used unless some pattern's constraint asks for one.
2222 @defmac HARD_REGNO_RENAME_OK (@var{from}, @var{to})
2223 A C expression that is nonzero if it is OK to rename a hard register
2224 @var{from} to another hard register @var{to}.
2226 One common use of this macro is to prevent renaming of a register to
2227 another register that is not saved by a prologue in an interrupt
2230 The default is always nonzero.
2233 @defmac MODES_TIEABLE_P (@var{mode1}, @var{mode2})
2234 A C expression that is nonzero if a value of mode
2235 @var{mode1} is accessible in mode @var{mode2} without copying.
2237 If @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode1})} and
2238 @code{HARD_REGNO_MODE_OK (@var{r}, @var{mode2})} are always the same for
2239 any @var{r}, then @code{MODES_TIEABLE_P (@var{mode1}, @var{mode2})}
2240 should be nonzero. If they differ for any @var{r}, you should define
2241 this macro to return zero unless some other mechanism ensures the
2242 accessibility of the value in a narrower mode.
2244 You should define this macro to return nonzero in as many cases as
2245 possible since doing so will allow GCC to perform better register
2249 @deftypefn {Target Hook} bool TARGET_HARD_REGNO_SCRATCH_OK (unsigned int @var{regno})
2250 This target hook should return @code{true} if it is OK to use a hard register
2251 @var{regno} as scratch reg in peephole2.
2253 One common use of this macro is to prevent using of a register that
2254 is not saved by a prologue in an interrupt handler.
2256 The default version of this hook always returns @code{true}.
2259 @defmac AVOID_CCMODE_COPIES
2260 Define this macro if the compiler should avoid copies to/from @code{CCmode}
2261 registers. You should only define this macro if support for copying to/from
2262 @code{CCmode} is incomplete.
2265 @node Leaf Functions
2266 @subsection Handling Leaf Functions
2268 @cindex leaf functions
2269 @cindex functions, leaf
2270 On some machines, a leaf function (i.e., one which makes no calls) can run
2271 more efficiently if it does not make its own register window. Often this
2272 means it is required to receive its arguments in the registers where they
2273 are passed by the caller, instead of the registers where they would
2276 The special treatment for leaf functions generally applies only when
2277 other conditions are met; for example, often they may use only those
2278 registers for its own variables and temporaries. We use the term ``leaf
2279 function'' to mean a function that is suitable for this special
2280 handling, so that functions with no calls are not necessarily ``leaf
2283 GCC assigns register numbers before it knows whether the function is
2284 suitable for leaf function treatment. So it needs to renumber the
2285 registers in order to output a leaf function. The following macros
2288 @defmac LEAF_REGISTERS
2289 Name of a char vector, indexed by hard register number, which
2290 contains 1 for a register that is allowable in a candidate for leaf
2293 If leaf function treatment involves renumbering the registers, then the
2294 registers marked here should be the ones before renumbering---those that
2295 GCC would ordinarily allocate. The registers which will actually be
2296 used in the assembler code, after renumbering, should not be marked with 1
2299 Define this macro only if the target machine offers a way to optimize
2300 the treatment of leaf functions.
2303 @defmac LEAF_REG_REMAP (@var{regno})
2304 A C expression whose value is the register number to which @var{regno}
2305 should be renumbered, when a function is treated as a leaf function.
2307 If @var{regno} is a register number which should not appear in a leaf
2308 function before renumbering, then the expression should yield @minus{}1, which
2309 will cause the compiler to abort.
2311 Define this macro only if the target machine offers a way to optimize the
2312 treatment of leaf functions, and registers need to be renumbered to do
2316 @findex current_function_is_leaf
2317 @findex current_function_uses_only_leaf_regs
2318 @code{TARGET_ASM_FUNCTION_PROLOGUE} and
2319 @code{TARGET_ASM_FUNCTION_EPILOGUE} must usually treat leaf functions
2320 specially. They can test the C variable @code{current_function_is_leaf}
2321 which is nonzero for leaf functions. @code{current_function_is_leaf} is
2322 set prior to local register allocation and is valid for the remaining
2323 compiler passes. They can also test the C variable
2324 @code{current_function_uses_only_leaf_regs} which is nonzero for leaf
2325 functions which only use leaf registers.
2326 @code{current_function_uses_only_leaf_regs} is valid after all passes
2327 that modify the instructions have been run and is only useful if
2328 @code{LEAF_REGISTERS} is defined.
2329 @c changed this to fix overfull. ALSO: why the "it" at the beginning
2330 @c of the next paragraph?! --mew 2feb93
2332 @node Stack Registers
2333 @subsection Registers That Form a Stack
2335 There are special features to handle computers where some of the
2336 ``registers'' form a stack. Stack registers are normally written by
2337 pushing onto the stack, and are numbered relative to the top of the
2340 Currently, GCC can only handle one group of stack-like registers, and
2341 they must be consecutively numbered. Furthermore, the existing
2342 support for stack-like registers is specific to the 80387 floating
2343 point coprocessor. If you have a new architecture that uses
2344 stack-like registers, you will need to do substantial work on
2345 @file{reg-stack.c} and write your machine description to cooperate
2346 with it, as well as defining these macros.
2349 Define this if the machine has any stack-like registers.
2352 @defmac STACK_REG_COVER_CLASS
2353 This is a cover class containing the stack registers. Define this if
2354 the machine has any stack-like registers.
2357 @defmac FIRST_STACK_REG
2358 The number of the first stack-like register. This one is the top
2362 @defmac LAST_STACK_REG
2363 The number of the last stack-like register. This one is the bottom of
2367 @node Register Classes
2368 @section Register Classes
2369 @cindex register class definitions
2370 @cindex class definitions, register
2372 On many machines, the numbered registers are not all equivalent.
2373 For example, certain registers may not be allowed for indexed addressing;
2374 certain registers may not be allowed in some instructions. These machine
2375 restrictions are described to the compiler using @dfn{register classes}.
2377 You define a number of register classes, giving each one a name and saying
2378 which of the registers belong to it. Then you can specify register classes
2379 that are allowed as operands to particular instruction patterns.
2383 In general, each register will belong to several classes. In fact, one
2384 class must be named @code{ALL_REGS} and contain all the registers. Another
2385 class must be named @code{NO_REGS} and contain no registers. Often the
2386 union of two classes will be another class; however, this is not required.
2388 @findex GENERAL_REGS
2389 One of the classes must be named @code{GENERAL_REGS}. There is nothing
2390 terribly special about the name, but the operand constraint letters
2391 @samp{r} and @samp{g} specify this class. If @code{GENERAL_REGS} is
2392 the same as @code{ALL_REGS}, just define it as a macro which expands
2395 Order the classes so that if class @var{x} is contained in class @var{y}
2396 then @var{x} has a lower class number than @var{y}.
2398 The way classes other than @code{GENERAL_REGS} are specified in operand
2399 constraints is through machine-dependent operand constraint letters.
2400 You can define such letters to correspond to various classes, then use
2401 them in operand constraints.
2403 You should define a class for the union of two classes whenever some
2404 instruction allows both classes. For example, if an instruction allows
2405 either a floating point (coprocessor) register or a general register for a
2406 certain operand, you should define a class @code{FLOAT_OR_GENERAL_REGS}
2407 which includes both of them. Otherwise you will get suboptimal code.
2409 You must also specify certain redundant information about the register
2410 classes: for each class, which classes contain it and which ones are
2411 contained in it; for each pair of classes, the largest class contained
2414 When a value occupying several consecutive registers is expected in a
2415 certain class, all the registers used must belong to that class.
2416 Therefore, register classes cannot be used to enforce a requirement for
2417 a register pair to start with an even-numbered register. The way to
2418 specify this requirement is with @code{HARD_REGNO_MODE_OK}.
2420 Register classes used for input-operands of bitwise-and or shift
2421 instructions have a special requirement: each such class must have, for
2422 each fixed-point machine mode, a subclass whose registers can transfer that
2423 mode to or from memory. For example, on some machines, the operations for
2424 single-byte values (@code{QImode}) are limited to certain registers. When
2425 this is so, each register class that is used in a bitwise-and or shift
2426 instruction must have a subclass consisting of registers from which
2427 single-byte values can be loaded or stored. This is so that
2428 @code{PREFERRED_RELOAD_CLASS} can always have a possible value to return.
2430 @deftp {Data type} {enum reg_class}
2431 An enumerated type that must be defined with all the register class names
2432 as enumerated values. @code{NO_REGS} must be first. @code{ALL_REGS}
2433 must be the last register class, followed by one more enumerated value,
2434 @code{LIM_REG_CLASSES}, which is not a register class but rather
2435 tells how many classes there are.
2437 Each register class has a number, which is the value of casting
2438 the class name to type @code{int}. The number serves as an index
2439 in many of the tables described below.
2442 @defmac N_REG_CLASSES
2443 The number of distinct register classes, defined as follows:
2446 #define N_REG_CLASSES (int) LIM_REG_CLASSES
2450 @defmac REG_CLASS_NAMES
2451 An initializer containing the names of the register classes as C string
2452 constants. These names are used in writing some of the debugging dumps.
2455 @defmac REG_CLASS_CONTENTS
2456 An initializer containing the contents of the register classes, as integers
2457 which are bit masks. The @var{n}th integer specifies the contents of class
2458 @var{n}. The way the integer @var{mask} is interpreted is that
2459 register @var{r} is in the class if @code{@var{mask} & (1 << @var{r})} is 1.
2461 When the machine has more than 32 registers, an integer does not suffice.
2462 Then the integers are replaced by sub-initializers, braced groupings containing
2463 several integers. Each sub-initializer must be suitable as an initializer
2464 for the type @code{HARD_REG_SET} which is defined in @file{hard-reg-set.h}.
2465 In this situation, the first integer in each sub-initializer corresponds to
2466 registers 0 through 31, the second integer to registers 32 through 63, and
2470 @defmac REGNO_REG_CLASS (@var{regno})
2471 A C expression whose value is a register class containing hard register
2472 @var{regno}. In general there is more than one such class; choose a class
2473 which is @dfn{minimal}, meaning that no smaller class also contains the
2477 @defmac BASE_REG_CLASS
2478 A macro whose definition is the name of the class to which a valid
2479 base register must belong. A base register is one used in an address
2480 which is the register value plus a displacement.
2483 @defmac MODE_BASE_REG_CLASS (@var{mode})
2484 This is a variation of the @code{BASE_REG_CLASS} macro which allows
2485 the selection of a base register in a mode dependent manner. If
2486 @var{mode} is VOIDmode then it should return the same value as
2487 @code{BASE_REG_CLASS}.
2490 @defmac MODE_BASE_REG_REG_CLASS (@var{mode})
2491 A C expression whose value is the register class to which a valid
2492 base register must belong in order to be used in a base plus index
2493 register address. You should define this macro if base plus index
2494 addresses have different requirements than other base register uses.
2497 @defmac MODE_CODE_BASE_REG_CLASS (@var{mode}, @var{outer_code}, @var{index_code})
2498 A C expression whose value is the register class to which a valid
2499 base register must belong. @var{outer_code} and @var{index_code} define the
2500 context in which the base register occurs. @var{outer_code} is the code of
2501 the immediately enclosing expression (@code{MEM} for the top level of an
2502 address, @code{ADDRESS} for something that occurs in an
2503 @code{address_operand}). @var{index_code} is the code of the corresponding
2504 index expression if @var{outer_code} is @code{PLUS}; @code{SCRATCH} otherwise.
2507 @defmac INDEX_REG_CLASS
2508 A macro whose definition is the name of the class to which a valid
2509 index register must belong. An index register is one used in an
2510 address where its value is either multiplied by a scale factor or
2511 added to another register (as well as added to a displacement).
2514 @defmac REGNO_OK_FOR_BASE_P (@var{num})
2515 A C expression which is nonzero if register number @var{num} is
2516 suitable for use as a base register in operand addresses.
2517 Like @code{TARGET_LEGITIMATE_ADDRESS_P}, this macro should also
2518 define a strict and a non-strict variant. Both variants behave
2519 the same for hard register; for pseudos, the strict variant will
2520 pass only those that have been allocated to a valid hard registers,
2521 while the non-strict variant will pass all pseudos.
2523 @findex REG_OK_STRICT
2524 Compiler source files that want to use the strict variant of this and
2525 other macros define the macro @code{REG_OK_STRICT}. You should use an
2526 @code{#ifdef REG_OK_STRICT} conditional to define the strict variant in
2527 that case and the non-strict variant otherwise.
2530 @defmac REGNO_MODE_OK_FOR_BASE_P (@var{num}, @var{mode})
2531 A C expression that is just like @code{REGNO_OK_FOR_BASE_P}, except that
2532 that expression may examine the mode of the memory reference in
2533 @var{mode}. You should define this macro if the mode of the memory
2534 reference affects whether a register may be used as a base register. If
2535 you define this macro, the compiler will use it instead of
2536 @code{REGNO_OK_FOR_BASE_P}. The mode may be @code{VOIDmode} for
2537 addresses that appear outside a @code{MEM}, i.e., as an
2538 @code{address_operand}.
2540 This macro also has strict and non-strict variants.
2543 @defmac REGNO_MODE_OK_FOR_REG_BASE_P (@var{num}, @var{mode})
2544 A C expression which is nonzero if register number @var{num} is suitable for
2545 use as a base register in base plus index operand addresses, accessing
2546 memory in mode @var{mode}. It may be either a suitable hard register or a
2547 pseudo register that has been allocated such a hard register. You should
2548 define this macro if base plus index addresses have different requirements
2549 than other base register uses.
2551 Use of this macro is deprecated; please use the more general
2552 @code{REGNO_MODE_CODE_OK_FOR_BASE_P}.
2554 This macro also has strict and non-strict variants.
2557 @defmac REGNO_MODE_CODE_OK_FOR_BASE_P (@var{num}, @var{mode}, @var{outer_code}, @var{index_code})
2558 A C expression that is just like @code{REGNO_MODE_OK_FOR_BASE_P}, except
2559 that that expression may examine the context in which the register
2560 appears in the memory reference. @var{outer_code} is the code of the
2561 immediately enclosing expression (@code{MEM} if at the top level of the
2562 address, @code{ADDRESS} for something that occurs in an
2563 @code{address_operand}). @var{index_code} is the code of the
2564 corresponding index expression if @var{outer_code} is @code{PLUS};
2565 @code{SCRATCH} otherwise. The mode may be @code{VOIDmode} for addresses
2566 that appear outside a @code{MEM}, i.e., as an @code{address_operand}.
2568 This macro also has strict and non-strict variants.
2571 @defmac REGNO_OK_FOR_INDEX_P (@var{num})
2572 A C expression which is nonzero if register number @var{num} is
2573 suitable for use as an index register in operand addresses. It may be
2574 either a suitable hard register or a pseudo register that has been
2575 allocated such a hard register.
2577 The difference between an index register and a base register is that
2578 the index register may be scaled. If an address involves the sum of
2579 two registers, neither one of them scaled, then either one may be
2580 labeled the ``base'' and the other the ``index''; but whichever
2581 labeling is used must fit the machine's constraints of which registers
2582 may serve in each capacity. The compiler will try both labelings,
2583 looking for one that is valid, and will reload one or both registers
2584 only if neither labeling works.
2586 This macro also has strict and non-strict variants.
2589 @defmac PREFERRED_RELOAD_CLASS (@var{x}, @var{class})
2590 A C expression that places additional restrictions on the register class
2591 to use when it is necessary to copy value @var{x} into a register in class
2592 @var{class}. The value is a register class; perhaps @var{class}, or perhaps
2593 another, smaller class. On many machines, the following definition is
2597 #define PREFERRED_RELOAD_CLASS(X,CLASS) CLASS
2600 Sometimes returning a more restrictive class makes better code. For
2601 example, on the 68000, when @var{x} is an integer constant that is in range
2602 for a @samp{moveq} instruction, the value of this macro is always
2603 @code{DATA_REGS} as long as @var{class} includes the data registers.
2604 Requiring a data register guarantees that a @samp{moveq} will be used.
2606 One case where @code{PREFERRED_RELOAD_CLASS} must not return
2607 @var{class} is if @var{x} is a legitimate constant which cannot be
2608 loaded into some register class. By returning @code{NO_REGS} you can
2609 force @var{x} into a memory location. For example, rs6000 can load
2610 immediate values into general-purpose registers, but does not have an
2611 instruction for loading an immediate value into a floating-point
2612 register, so @code{PREFERRED_RELOAD_CLASS} returns @code{NO_REGS} when
2613 @var{x} is a floating-point constant. If the constant can't be loaded
2614 into any kind of register, code generation will be better if
2615 @code{LEGITIMATE_CONSTANT_P} makes the constant illegitimate instead
2616 of using @code{PREFERRED_RELOAD_CLASS}.
2618 If an insn has pseudos in it after register allocation, reload will go
2619 through the alternatives and call repeatedly @code{PREFERRED_RELOAD_CLASS}
2620 to find the best one. Returning @code{NO_REGS}, in this case, makes
2621 reload add a @code{!} in front of the constraint: the x86 back-end uses
2622 this feature to discourage usage of 387 registers when math is done in
2623 the SSE registers (and vice versa).
2626 @defmac PREFERRED_OUTPUT_RELOAD_CLASS (@var{x}, @var{class})
2627 Like @code{PREFERRED_RELOAD_CLASS}, but for output reloads instead of
2628 input reloads. If you don't define this macro, the default is to use
2629 @var{class}, unchanged.
2631 You can also use @code{PREFERRED_OUTPUT_RELOAD_CLASS} to discourage
2632 reload from using some alternatives, like @code{PREFERRED_RELOAD_CLASS}.
2635 @defmac LIMIT_RELOAD_CLASS (@var{mode}, @var{class})
2636 A C expression that places additional restrictions on the register class
2637 to use when it is necessary to be able to hold a value of mode
2638 @var{mode} in a reload register for which class @var{class} would
2641 Unlike @code{PREFERRED_RELOAD_CLASS}, this macro should be used when
2642 there are certain modes that simply can't go in certain reload classes.
2644 The value is a register class; perhaps @var{class}, or perhaps another,
2647 Don't define this macro unless the target machine has limitations which
2648 require the macro to do something nontrivial.
2651 @deftypefn {Target Hook} {enum reg_class} TARGET_SECONDARY_RELOAD (bool @var{in_p}, rtx @var{x}, enum reg_class @var{reload_class}, enum machine_mode @var{reload_mode}, secondary_reload_info *@var{sri})
2652 Many machines have some registers that cannot be copied directly to or
2653 from memory or even from other types of registers. An example is the
2654 @samp{MQ} register, which on most machines, can only be copied to or
2655 from general registers, but not memory. Below, we shall be using the
2656 term 'intermediate register' when a move operation cannot be performed
2657 directly, but has to be done by copying the source into the intermediate
2658 register first, and then copying the intermediate register to the
2659 destination. An intermediate register always has the same mode as
2660 source and destination. Since it holds the actual value being copied,
2661 reload might apply optimizations to re-use an intermediate register
2662 and eliding the copy from the source when it can determine that the
2663 intermediate register still holds the required value.
2665 Another kind of secondary reload is required on some machines which
2666 allow copying all registers to and from memory, but require a scratch
2667 register for stores to some memory locations (e.g., those with symbolic
2668 address on the RT, and those with certain symbolic address on the SPARC
2669 when compiling PIC)@. Scratch registers need not have the same mode
2670 as the value being copied, and usually hold a different value than
2671 that being copied. Special patterns in the md file are needed to
2672 describe how the copy is performed with the help of the scratch register;
2673 these patterns also describe the number, register class(es) and mode(s)
2674 of the scratch register(s).
2676 In some cases, both an intermediate and a scratch register are required.
2678 For input reloads, this target hook is called with nonzero @var{in_p},
2679 and @var{x} is an rtx that needs to be copied to a register of class
2680 @var{reload_class} in @var{reload_mode}. For output reloads, this target
2681 hook is called with zero @var{in_p}, and a register of class @var{reload_class}
2682 needs to be copied to rtx @var{x} in @var{reload_mode}.
2684 If copying a register of @var{reload_class} from/to @var{x} requires
2685 an intermediate register, the hook @code{secondary_reload} should
2686 return the register class required for this intermediate register.
2687 If no intermediate register is required, it should return NO_REGS.
2688 If more than one intermediate register is required, describe the one
2689 that is closest in the copy chain to the reload register.
2691 If scratch registers are needed, you also have to describe how to
2692 perform the copy from/to the reload register to/from this
2693 closest intermediate register. Or if no intermediate register is
2694 required, but still a scratch register is needed, describe the
2695 copy from/to the reload register to/from the reload operand @var{x}.
2697 You do this by setting @code{sri->icode} to the instruction code of a pattern
2698 in the md file which performs the move. Operands 0 and 1 are the output
2699 and input of this copy, respectively. Operands from operand 2 onward are
2700 for scratch operands. These scratch operands must have a mode, and a
2701 single-register-class
2702 @c [later: or memory]
2705 When an intermediate register is used, the @code{secondary_reload}
2706 hook will be called again to determine how to copy the intermediate
2707 register to/from the reload operand @var{x}, so your hook must also
2708 have code to handle the register class of the intermediate operand.
2710 @c [For later: maybe we'll allow multi-alternative reload patterns -
2711 @c the port maintainer could name a mov<mode> pattern that has clobbers -
2712 @c and match the constraints of input and output to determine the required
2713 @c alternative. A restriction would be that constraints used to match
2714 @c against reloads registers would have to be written as register class
2715 @c constraints, or we need a new target macro / hook that tells us if an
2716 @c arbitrary constraint can match an unknown register of a given class.
2717 @c Such a macro / hook would also be useful in other places.]
2720 @var{x} might be a pseudo-register or a @code{subreg} of a
2721 pseudo-register, which could either be in a hard register or in memory.
2722 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2723 in memory and the hard register number if it is in a register.
2725 Scratch operands in memory (constraint @code{"=m"} / @code{"=&m"}) are
2726 currently not supported. For the time being, you will have to continue
2727 to use @code{SECONDARY_MEMORY_NEEDED} for that purpose.
2729 @code{copy_cost} also uses this target hook to find out how values are
2730 copied. If you want it to include some extra cost for the need to allocate
2731 (a) scratch register(s), set @code{sri->extra_cost} to the additional cost.
2732 Or if two dependent moves are supposed to have a lower cost than the sum
2733 of the individual moves due to expected fortuitous scheduling and/or special
2734 forwarding logic, you can set @code{sri->extra_cost} to a negative amount.
2737 @defmac SECONDARY_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2738 @defmacx SECONDARY_INPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2739 @defmacx SECONDARY_OUTPUT_RELOAD_CLASS (@var{class}, @var{mode}, @var{x})
2740 These macros are obsolete, new ports should use the target hook
2741 @code{TARGET_SECONDARY_RELOAD} instead.
2743 These are obsolete macros, replaced by the @code{TARGET_SECONDARY_RELOAD}
2744 target hook. Older ports still define these macros to indicate to the
2745 reload phase that it may
2746 need to allocate at least one register for a reload in addition to the
2747 register to contain the data. Specifically, if copying @var{x} to a
2748 register @var{class} in @var{mode} requires an intermediate register,
2749 you were supposed to define @code{SECONDARY_INPUT_RELOAD_CLASS} to return the
2750 largest register class all of whose registers can be used as
2751 intermediate registers or scratch registers.
2753 If copying a register @var{class} in @var{mode} to @var{x} requires an
2754 intermediate or scratch register, @code{SECONDARY_OUTPUT_RELOAD_CLASS}
2755 was supposed to be defined be defined to return the largest register
2756 class required. If the
2757 requirements for input and output reloads were the same, the macro
2758 @code{SECONDARY_RELOAD_CLASS} should have been used instead of defining both
2761 The values returned by these macros are often @code{GENERAL_REGS}.
2762 Return @code{NO_REGS} if no spare register is needed; i.e., if @var{x}
2763 can be directly copied to or from a register of @var{class} in
2764 @var{mode} without requiring a scratch register. Do not define this
2765 macro if it would always return @code{NO_REGS}.
2767 If a scratch register is required (either with or without an
2768 intermediate register), you were supposed to define patterns for
2769 @samp{reload_in@var{m}} or @samp{reload_out@var{m}}, as required
2770 (@pxref{Standard Names}. These patterns, which were normally
2771 implemented with a @code{define_expand}, should be similar to the
2772 @samp{mov@var{m}} patterns, except that operand 2 is the scratch
2775 These patterns need constraints for the reload register and scratch
2777 contain a single register class. If the original reload register (whose
2778 class is @var{class}) can meet the constraint given in the pattern, the
2779 value returned by these macros is used for the class of the scratch
2780 register. Otherwise, two additional reload registers are required.
2781 Their classes are obtained from the constraints in the insn pattern.
2783 @var{x} might be a pseudo-register or a @code{subreg} of a
2784 pseudo-register, which could either be in a hard register or in memory.
2785 Use @code{true_regnum} to find out; it will return @minus{}1 if the pseudo is
2786 in memory and the hard register number if it is in a register.
2788 These macros should not be used in the case where a particular class of
2789 registers can only be copied to memory and not to another class of
2790 registers. In that case, secondary reload registers are not needed and
2791 would not be helpful. Instead, a stack location must be used to perform
2792 the copy and the @code{mov@var{m}} pattern should use memory as an
2793 intermediate storage. This case often occurs between floating-point and
2797 @defmac SECONDARY_MEMORY_NEEDED (@var{class1}, @var{class2}, @var{m})
2798 Certain machines have the property that some registers cannot be copied
2799 to some other registers without using memory. Define this macro on
2800 those machines to be a C expression that is nonzero if objects of mode
2801 @var{m} in registers of @var{class1} can only be copied to registers of
2802 class @var{class2} by storing a register of @var{class1} into memory
2803 and loading that memory location into a register of @var{class2}.
2805 Do not define this macro if its value would always be zero.
2808 @defmac SECONDARY_MEMORY_NEEDED_RTX (@var{mode})
2809 Normally when @code{SECONDARY_MEMORY_NEEDED} is defined, the compiler
2810 allocates a stack slot for a memory location needed for register copies.
2811 If this macro is defined, the compiler instead uses the memory location
2812 defined by this macro.
2814 Do not define this macro if you do not define
2815 @code{SECONDARY_MEMORY_NEEDED}.
2818 @defmac SECONDARY_MEMORY_NEEDED_MODE (@var{mode})
2819 When the compiler needs a secondary memory location to copy between two
2820 registers of mode @var{mode}, it normally allocates sufficient memory to
2821 hold a quantity of @code{BITS_PER_WORD} bits and performs the store and
2822 load operations in a mode that many bits wide and whose class is the
2823 same as that of @var{mode}.
2825 This is right thing to do on most machines because it ensures that all
2826 bits of the register are copied and prevents accesses to the registers
2827 in a narrower mode, which some machines prohibit for floating-point
2830 However, this default behavior is not correct on some machines, such as
2831 the DEC Alpha, that store short integers in floating-point registers
2832 differently than in integer registers. On those machines, the default
2833 widening will not work correctly and you must define this macro to
2834 suppress that widening in some cases. See the file @file{alpha.h} for
2837 Do not define this macro if you do not define
2838 @code{SECONDARY_MEMORY_NEEDED} or if widening @var{mode} to a mode that
2839 is @code{BITS_PER_WORD} bits wide is correct for your machine.
2842 @defmac SMALL_REGISTER_CLASSES
2843 On some machines, it is risky to let hard registers live across arbitrary
2844 insns. Typically, these machines have instructions that require values
2845 to be in specific registers (like an accumulator), and reload will fail
2846 if the required hard register is used for another purpose across such an
2849 Define @code{SMALL_REGISTER_CLASSES} to be an expression with a nonzero
2850 value on these machines. When this macro has a nonzero value, the
2851 compiler will try to minimize the lifetime of hard registers.
2853 It is always safe to define this macro with a nonzero value, but if you
2854 unnecessarily define it, you will reduce the amount of optimizations
2855 that can be performed in some cases. If you do not define this macro
2856 with a nonzero value when it is required, the compiler will run out of
2857 spill registers and print a fatal error message. For most machines, you
2858 should not define this macro at all.
2861 @defmac CLASS_LIKELY_SPILLED_P (@var{class})
2862 A C expression whose value is nonzero if pseudos that have been assigned
2863 to registers of class @var{class} would likely be spilled because
2864 registers of @var{class} are needed for spill registers.
2866 The default value of this macro returns 1 if @var{class} has exactly one
2867 register and zero otherwise. On most machines, this default should be
2868 used. Only define this macro to some other expression if pseudos
2869 allocated by @file{local-alloc.c} end up in memory because their hard
2870 registers were needed for spill registers. If this macro returns nonzero
2871 for those classes, those pseudos will only be allocated by
2872 @file{global.c}, which knows how to reallocate the pseudo to another
2873 register. If there would not be another register available for
2874 reallocation, you should not change the definition of this macro since
2875 the only effect of such a definition would be to slow down register
2879 @defmac CLASS_MAX_NREGS (@var{class}, @var{mode})
2880 A C expression for the maximum number of consecutive registers
2881 of class @var{class} needed to hold a value of mode @var{mode}.
2883 This is closely related to the macro @code{HARD_REGNO_NREGS}. In fact,
2884 the value of the macro @code{CLASS_MAX_NREGS (@var{class}, @var{mode})}
2885 should be the maximum value of @code{HARD_REGNO_NREGS (@var{regno},
2886 @var{mode})} for all @var{regno} values in the class @var{class}.
2888 This macro helps control the handling of multiple-word values
2892 @defmac CANNOT_CHANGE_MODE_CLASS (@var{from}, @var{to}, @var{class})
2893 If defined, a C expression that returns nonzero for a @var{class} for which
2894 a change from mode @var{from} to mode @var{to} is invalid.
2896 For the example, loading 32-bit integer or floating-point objects into
2897 floating-point registers on the Alpha extends them to 64 bits.
2898 Therefore loading a 64-bit object and then storing it as a 32-bit object
2899 does not store the low-order 32 bits, as would be the case for a normal
2900 register. Therefore, @file{alpha.h} defines @code{CANNOT_CHANGE_MODE_CLASS}
2904 #define CANNOT_CHANGE_MODE_CLASS(FROM, TO, CLASS) \
2905 (GET_MODE_SIZE (FROM) != GET_MODE_SIZE (TO) \
2906 ? reg_classes_intersect_p (FLOAT_REGS, (CLASS)) : 0)
2910 @deftypefn {Target Hook} {const enum reg_class *} TARGET_IRA_COVER_CLASSES ()
2911 Return an array of cover classes for the Integrated Register Allocator
2912 (@acronym{IRA}). Cover classes are a set of non-intersecting register
2913 classes covering all hard registers used for register allocation
2914 purposes. If a move between two registers in the same cover class is
2915 possible, it should be cheaper than a load or store of the registers.
2916 The array is terminated by a @code{LIM_REG_CLASSES} element.
2918 The order of cover classes in the array is important. If two classes
2919 have the same cost of usage for a pseudo, the class occurred first in
2920 the array is chosen for the pseudo.
2922 This hook is called once at compiler startup, after the command-line
2923 options have been processed. It is then re-examined by every call to
2924 @code{target_reinit}.
2926 The default implementation returns @code{IRA_COVER_CLASSES}, if defined,
2927 otherwise there is no default implementation. You must define either this
2928 macro or @code{IRA_COVER_CLASSES} in order to use the integrated register
2929 allocator with Chaitin-Briggs coloring. If the macro is not defined,
2930 the only available coloring algorithm is Chow's priority coloring.
2933 @defmac IRA_COVER_CLASSES
2934 See the documentation for @code{TARGET_IRA_COVER_CLASSES}.
2937 @node Old Constraints
2938 @section Obsolete Macros for Defining Constraints
2939 @cindex defining constraints, obsolete method
2940 @cindex constraints, defining, obsolete method
2942 Machine-specific constraints can be defined with these macros instead
2943 of the machine description constructs described in @ref{Define
2944 Constraints}. This mechanism is obsolete. New ports should not use
2945 it; old ports should convert to the new mechanism.
2947 @defmac CONSTRAINT_LEN (@var{char}, @var{str})
2948 For the constraint at the start of @var{str}, which starts with the letter
2949 @var{c}, return the length. This allows you to have register class /
2950 constant / extra constraints that are longer than a single letter;
2951 you don't need to define this macro if you can do with single-letter
2952 constraints only. The definition of this macro should use
2953 DEFAULT_CONSTRAINT_LEN for all the characters that you don't want
2954 to handle specially.
2955 There are some sanity checks in genoutput.c that check the constraint lengths
2956 for the md file, so you can also use this macro to help you while you are
2957 transitioning from a byzantine single-letter-constraint scheme: when you
2958 return a negative length for a constraint you want to re-use, genoutput
2959 will complain about every instance where it is used in the md file.
2962 @defmac REG_CLASS_FROM_LETTER (@var{char})
2963 A C expression which defines the machine-dependent operand constraint
2964 letters for register classes. If @var{char} is such a letter, the
2965 value should be the register class corresponding to it. Otherwise,
2966 the value should be @code{NO_REGS}. The register letter @samp{r},
2967 corresponding to class @code{GENERAL_REGS}, will not be passed
2968 to this macro; you do not need to handle it.
2971 @defmac REG_CLASS_FROM_CONSTRAINT (@var{char}, @var{str})
2972 Like @code{REG_CLASS_FROM_LETTER}, but you also get the constraint string
2973 passed in @var{str}, so that you can use suffixes to distinguish between
2977 @defmac CONST_OK_FOR_LETTER_P (@var{value}, @var{c})
2978 A C expression that defines the machine-dependent operand constraint
2979 letters (@samp{I}, @samp{J}, @samp{K}, @dots{} @samp{P}) that specify
2980 particular ranges of integer values. If @var{c} is one of those
2981 letters, the expression should check that @var{value}, an integer, is in
2982 the appropriate range and return 1 if so, 0 otherwise. If @var{c} is
2983 not one of those letters, the value should be 0 regardless of
2987 @defmac CONST_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
2988 Like @code{CONST_OK_FOR_LETTER_P}, but you also get the constraint
2989 string passed in @var{str}, so that you can use suffixes to distinguish
2990 between different variants.
2993 @defmac CONST_DOUBLE_OK_FOR_LETTER_P (@var{value}, @var{c})
2994 A C expression that defines the machine-dependent operand constraint
2995 letters that specify particular ranges of @code{const_double} values
2996 (@samp{G} or @samp{H}).
2998 If @var{c} is one of those letters, the expression should check that
2999 @var{value}, an RTX of code @code{const_double}, is in the appropriate
3000 range and return 1 if so, 0 otherwise. If @var{c} is not one of those
3001 letters, the value should be 0 regardless of @var{value}.
3003 @code{const_double} is used for all floating-point constants and for
3004 @code{DImode} fixed-point constants. A given letter can accept either
3005 or both kinds of values. It can use @code{GET_MODE} to distinguish
3006 between these kinds.
3009 @defmac CONST_DOUBLE_OK_FOR_CONSTRAINT_P (@var{value}, @var{c}, @var{str})
3010 Like @code{CONST_DOUBLE_OK_FOR_LETTER_P}, but you also get the constraint
3011 string passed in @var{str}, so that you can use suffixes to distinguish
3012 between different variants.
3015 @defmac EXTRA_CONSTRAINT (@var{value}, @var{c})
3016 A C expression that defines the optional machine-dependent constraint
3017 letters that can be used to segregate specific types of operands, usually
3018 memory references, for the target machine. Any letter that is not
3019 elsewhere defined and not matched by @code{REG_CLASS_FROM_LETTER} /
3020 @code{REG_CLASS_FROM_CONSTRAINT}
3021 may be used. Normally this macro will not be defined.
3023 If it is required for a particular target machine, it should return 1
3024 if @var{value} corresponds to the operand type represented by the
3025 constraint letter @var{c}. If @var{c} is not defined as an extra
3026 constraint, the value returned should be 0 regardless of @var{value}.
3028 For example, on the ROMP, load instructions cannot have their output
3029 in r0 if the memory reference contains a symbolic address. Constraint
3030 letter @samp{Q} is defined as representing a memory address that does
3031 @emph{not} contain a symbolic address. An alternative is specified with
3032 a @samp{Q} constraint on the input and @samp{r} on the output. The next
3033 alternative specifies @samp{m} on the input and a register class that
3034 does not include r0 on the output.
3037 @defmac EXTRA_CONSTRAINT_STR (@var{value}, @var{c}, @var{str})
3038 Like @code{EXTRA_CONSTRAINT}, but you also get the constraint string passed
3039 in @var{str}, so that you can use suffixes to distinguish between different
3043 @defmac EXTRA_MEMORY_CONSTRAINT (@var{c}, @var{str})
3044 A C expression that defines the optional machine-dependent constraint
3045 letters, amongst those accepted by @code{EXTRA_CONSTRAINT}, that should
3046 be treated like memory constraints by the reload pass.
3048 It should return 1 if the operand type represented by the constraint
3049 at the start of @var{str}, the first letter of which is the letter @var{c},
3050 comprises a subset of all memory references including
3051 all those whose address is simply a base register. This allows the reload
3052 pass to reload an operand, if it does not directly correspond to the operand
3053 type of @var{c}, by copying its address into a base register.
3055 For example, on the S/390, some instructions do not accept arbitrary
3056 memory references, but only those that do not make use of an index
3057 register. The constraint letter @samp{Q} is defined via
3058 @code{EXTRA_CONSTRAINT} as representing a memory address of this type.
3059 If the letter @samp{Q} is marked as @code{EXTRA_MEMORY_CONSTRAINT},
3060 a @samp{Q} constraint can handle any memory operand, because the
3061 reload pass knows it can be reloaded by copying the memory address
3062 into a base register if required. This is analogous to the way
3063 an @samp{o} constraint can handle any memory operand.
3066 @defmac EXTRA_ADDRESS_CONSTRAINT (@var{c}, @var{str})
3067 A C expression that defines the optional machine-dependent constraint
3068 letters, amongst those accepted by @code{EXTRA_CONSTRAINT} /
3069 @code{EXTRA_CONSTRAINT_STR}, that should
3070 be treated like address constraints by the reload pass.
3072 It should return 1 if the operand type represented by the constraint
3073 at the start of @var{str}, which starts with the letter @var{c}, comprises
3074 a subset of all memory addresses including
3075 all those that consist of just a base register. This allows the reload
3076 pass to reload an operand, if it does not directly correspond to the operand
3077 type of @var{str}, by copying it into a base register.
3079 Any constraint marked as @code{EXTRA_ADDRESS_CONSTRAINT} can only
3080 be used with the @code{address_operand} predicate. It is treated
3081 analogously to the @samp{p} constraint.
3084 @node Stack and Calling
3085 @section Stack Layout and Calling Conventions
3086 @cindex calling conventions
3088 @c prevent bad page break with this line
3089 This describes the stack layout and calling conventions.
3093 * Exception Handling::
3098 * Register Arguments::
3100 * Aggregate Return::
3105 * Stack Smashing Protection::
3109 @subsection Basic Stack Layout
3110 @cindex stack frame layout
3111 @cindex frame layout
3113 @c prevent bad page break with this line
3114 Here is the basic stack layout.
3116 @defmac STACK_GROWS_DOWNWARD
3117 Define this macro if pushing a word onto the stack moves the stack
3118 pointer to a smaller address.
3120 When we say, ``define this macro if @dots{}'', it means that the
3121 compiler checks this macro only with @code{#ifdef} so the precise
3122 definition used does not matter.
3125 @defmac STACK_PUSH_CODE
3126 This macro defines the operation used when something is pushed
3127 on the stack. In RTL, a push operation will be
3128 @code{(set (mem (STACK_PUSH_CODE (reg sp))) @dots{})}
3130 The choices are @code{PRE_DEC}, @code{POST_DEC}, @code{PRE_INC},
3131 and @code{POST_INC}. Which of these is correct depends on
3132 the stack direction and on whether the stack pointer points
3133 to the last item on the stack or whether it points to the
3134 space for the next item on the stack.
3136 The default is @code{PRE_DEC} when @code{STACK_GROWS_DOWNWARD} is
3137 defined, which is almost always right, and @code{PRE_INC} otherwise,
3138 which is often wrong.
3141 @defmac FRAME_GROWS_DOWNWARD
3142 Define this macro to nonzero value if the addresses of local variable slots
3143 are at negative offsets from the frame pointer.
3146 @defmac ARGS_GROW_DOWNWARD
3147 Define this macro if successive arguments to a function occupy decreasing
3148 addresses on the stack.
3151 @defmac STARTING_FRAME_OFFSET
3152 Offset from the frame pointer to the first local variable slot to be allocated.
3154 If @code{FRAME_GROWS_DOWNWARD}, find the next slot's offset by
3155 subtracting the first slot's length from @code{STARTING_FRAME_OFFSET}.
3156 Otherwise, it is found by adding the length of the first slot to the
3157 value @code{STARTING_FRAME_OFFSET}.
3158 @c i'm not sure if the above is still correct.. had to change it to get
3159 @c rid of an overfull. --mew 2feb93
3162 @defmac STACK_ALIGNMENT_NEEDED
3163 Define to zero to disable final alignment of the stack during reload.
3164 The nonzero default for this macro is suitable for most ports.
3166 On ports where @code{STARTING_FRAME_OFFSET} is nonzero or where there
3167 is a register save block following the local block that doesn't require
3168 alignment to @code{STACK_BOUNDARY}, it may be beneficial to disable
3169 stack alignment and do it in the backend.
3172 @defmac STACK_POINTER_OFFSET
3173 Offset from the stack pointer register to the first location at which
3174 outgoing arguments are placed. If not specified, the default value of
3175 zero is used. This is the proper value for most machines.
3177 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3178 the first location at which outgoing arguments are placed.
3181 @defmac FIRST_PARM_OFFSET (@var{fundecl})
3182 Offset from the argument pointer register to the first argument's
3183 address. On some machines it may depend on the data type of the
3186 If @code{ARGS_GROW_DOWNWARD}, this is the offset to the location above
3187 the first argument's address.
3190 @defmac STACK_DYNAMIC_OFFSET (@var{fundecl})
3191 Offset from the stack pointer register to an item dynamically allocated
3192 on the stack, e.g., by @code{alloca}.
3194 The default value for this macro is @code{STACK_POINTER_OFFSET} plus the
3195 length of the outgoing arguments. The default is correct for most
3196 machines. See @file{function.c} for details.
3199 @defmac INITIAL_FRAME_ADDRESS_RTX
3200 A C expression whose value is RTL representing the address of the initial
3201 stack frame. This address is passed to @code{RETURN_ADDR_RTX} and
3202 @code{DYNAMIC_CHAIN_ADDRESS}. If you don't define this macro, a reasonable
3203 default value will be used. Define this macro in order to make frame pointer
3204 elimination work in the presence of @code{__builtin_frame_address (count)} and
3205 @code{__builtin_return_address (count)} for @code{count} not equal to zero.
3208 @defmac DYNAMIC_CHAIN_ADDRESS (@var{frameaddr})
3209 A C expression whose value is RTL representing the address in a stack
3210 frame where the pointer to the caller's frame is stored. Assume that
3211 @var{frameaddr} is an RTL expression for the address of the stack frame
3214 If you don't define this macro, the default is to return the value
3215 of @var{frameaddr}---that is, the stack frame address is also the
3216 address of the stack word that points to the previous frame.
3219 @defmac SETUP_FRAME_ADDRESSES
3220 If defined, a C expression that produces the machine-specific code to
3221 setup the stack so that arbitrary frames can be accessed. For example,
3222 on the SPARC, we must flush all of the register windows to the stack
3223 before we can access arbitrary stack frames. You will seldom need to
3227 @deftypefn {Target Hook} bool TARGET_BUILTIN_SETJMP_FRAME_VALUE ()
3228 This target hook should return an rtx that is used to store
3229 the address of the current frame into the built in @code{setjmp} buffer.
3230 The default value, @code{virtual_stack_vars_rtx}, is correct for most
3231 machines. One reason you may need to define this target hook is if
3232 @code{hard_frame_pointer_rtx} is the appropriate value on your machine.
3235 @defmac FRAME_ADDR_RTX (@var{frameaddr})
3236 A C expression whose value is RTL representing the value of the frame
3237 address for the current frame. @var{frameaddr} is the frame pointer
3238 of the current frame. This is used for __builtin_frame_address.
3239 You need only define this macro if the frame address is not the same
3240 as the frame pointer. Most machines do not need to define it.
3243 @defmac RETURN_ADDR_RTX (@var{count}, @var{frameaddr})
3244 A C expression whose value is RTL representing the value of the return
3245 address for the frame @var{count} steps up from the current frame, after
3246 the prologue. @var{frameaddr} is the frame pointer of the @var{count}
3247 frame, or the frame pointer of the @var{count} @minus{} 1 frame if
3248 @code{RETURN_ADDR_IN_PREVIOUS_FRAME} is defined.
3250 The value of the expression must always be the correct address when
3251 @var{count} is zero, but may be @code{NULL_RTX} if there is no way to
3252 determine the return address of other frames.
3255 @defmac RETURN_ADDR_IN_PREVIOUS_FRAME
3256 Define this if the return address of a particular stack frame is accessed
3257 from the frame pointer of the previous stack frame.
3260 @defmac INCOMING_RETURN_ADDR_RTX
3261 A C expression whose value is RTL representing the location of the
3262 incoming return address at the beginning of any function, before the
3263 prologue. This RTL is either a @code{REG}, indicating that the return
3264 value is saved in @samp{REG}, or a @code{MEM} representing a location in
3267 You only need to define this macro if you want to support call frame
3268 debugging information like that provided by DWARF 2.
3270 If this RTL is a @code{REG}, you should also define
3271 @code{DWARF_FRAME_RETURN_COLUMN} to @code{DWARF_FRAME_REGNUM (REGNO)}.
3274 @defmac DWARF_ALT_FRAME_RETURN_COLUMN
3275 A C expression whose value is an integer giving a DWARF 2 column
3276 number that may be used as an alternative return column. The column
3277 must not correspond to any gcc hard register (that is, it must not
3278 be in the range of @code{DWARF_FRAME_REGNUM}).
3280 This macro can be useful if @code{DWARF_FRAME_RETURN_COLUMN} is set to a
3281 general register, but an alternative column needs to be used for signal
3282 frames. Some targets have also used different frame return columns
3286 @defmac DWARF_ZERO_REG
3287 A C expression whose value is an integer giving a DWARF 2 register
3288 number that is considered to always have the value zero. This should
3289 only be defined if the target has an architected zero register, and
3290 someone decided it was a good idea to use that register number to
3291 terminate the stack backtrace. New ports should avoid this.
3294 @deftypefn {Target Hook} void TARGET_DWARF_HANDLE_FRAME_UNSPEC (const char *@var{label}, rtx @var{pattern}, int @var{index})
3295 This target hook allows the backend to emit frame-related insns that
3296 contain UNSPECs or UNSPEC_VOLATILEs. The DWARF 2 call frame debugging
3297 info engine will invoke it on insns of the form
3299 (set (reg) (unspec [@dots{}] UNSPEC_INDEX))
3303 (set (reg) (unspec_volatile [@dots{}] UNSPECV_INDEX)).
3305 to let the backend emit the call frame instructions. @var{label} is
3306 the CFI label attached to the insn, @var{pattern} is the pattern of
3307 the insn and @var{index} is @code{UNSPEC_INDEX} or @code{UNSPECV_INDEX}.
3310 @defmac INCOMING_FRAME_SP_OFFSET
3311 A C expression whose value is an integer giving the offset, in bytes,
3312 from the value of the stack pointer register to the top of the stack
3313 frame at the beginning of any function, before the prologue. The top of
3314 the frame is defined to be the value of the stack pointer in the
3315 previous frame, just before the call instruction.
3317 You only need to define this macro if you want to support call frame
3318 debugging information like that provided by DWARF 2.
3321 @defmac ARG_POINTER_CFA_OFFSET (@var{fundecl})
3322 A C expression whose value is an integer giving the offset, in bytes,
3323 from the argument pointer to the canonical frame address (cfa). The
3324 final value should coincide with that calculated by
3325 @code{INCOMING_FRAME_SP_OFFSET}. Which is unfortunately not usable
3326 during virtual register instantiation.
3328 The default value for this macro is @code{FIRST_PARM_OFFSET (fundecl)},
3329 which is correct for most machines; in general, the arguments are found
3330 immediately before the stack frame. Note that this is not the case on
3331 some targets that save registers into the caller's frame, such as SPARC
3332 and rs6000, and so such targets need to define this macro.
3334 You only need to define this macro if the default is incorrect, and you
3335 want to support call frame debugging information like that provided by
3339 @defmac FRAME_POINTER_CFA_OFFSET (@var{fundecl})
3340 If defined, a C expression whose value is an integer giving the offset
3341 in bytes from the frame pointer to the canonical frame address (cfa).
3342 The final value should coincide with that calculated by
3343 @code{INCOMING_FRAME_SP_OFFSET}.
3345 Normally the CFA is calculated as an offset from the argument pointer,
3346 via @code{ARG_POINTER_CFA_OFFSET}, but if the argument pointer is
3347 variable due to the ABI, this may not be possible. If this macro is
3348 defined, it implies that the virtual register instantiation should be
3349 based on the frame pointer instead of the argument pointer. Only one
3350 of @code{FRAME_POINTER_CFA_OFFSET} and @code{ARG_POINTER_CFA_OFFSET}
3354 @defmac CFA_FRAME_BASE_OFFSET (@var{fundecl})
3355 If defined, a C expression whose value is an integer giving the offset
3356 in bytes from the canonical frame address (cfa) to the frame base used
3357 in DWARF 2 debug information. The default is zero. A different value
3358 may reduce the size of debug information on some ports.
3361 @node Exception Handling
3362 @subsection Exception Handling Support
3363 @cindex exception handling
3365 @defmac EH_RETURN_DATA_REGNO (@var{N})
3366 A C expression whose value is the @var{N}th register number used for
3367 data by exception handlers, or @code{INVALID_REGNUM} if fewer than
3368 @var{N} registers are usable.
3370 The exception handling library routines communicate with the exception
3371 handlers via a set of agreed upon registers. Ideally these registers
3372 should be call-clobbered; it is possible to use call-saved registers,
3373 but may negatively impact code size. The target must support at least
3374 2 data registers, but should define 4 if there are enough free registers.
3376 You must define this macro if you want to support call frame exception
3377 handling like that provided by DWARF 2.
3380 @defmac EH_RETURN_STACKADJ_RTX
3381 A C expression whose value is RTL representing a location in which
3382 to store a stack adjustment to be applied before function return.
3383 This is used to unwind the stack to an exception handler's call frame.
3384 It will be assigned zero on code paths that return normally.
3386 Typically this is a call-clobbered hard register that is otherwise
3387 untouched by the epilogue, but could also be a stack slot.
3389 Do not define this macro if the stack pointer is saved and restored
3390 by the regular prolog and epilog code in the call frame itself; in
3391 this case, the exception handling library routines will update the
3392 stack location to be restored in place. Otherwise, you must define
3393 this macro if you want to support call frame exception handling like
3394 that provided by DWARF 2.
3397 @defmac EH_RETURN_HANDLER_RTX
3398 A C expression whose value is RTL representing a location in which
3399 to store the address of an exception handler to which we should
3400 return. It will not be assigned on code paths that return normally.
3402 Typically this is the location in the call frame at which the normal
3403 return address is stored. For targets that return by popping an
3404 address off the stack, this might be a memory address just below
3405 the @emph{target} call frame rather than inside the current call
3406 frame. If defined, @code{EH_RETURN_STACKADJ_RTX} will have already
3407 been assigned, so it may be used to calculate the location of the
3410 Some targets have more complex requirements than storing to an
3411 address calculable during initial code generation. In that case
3412 the @code{eh_return} instruction pattern should be used instead.
3414 If you want to support call frame exception handling, you must
3415 define either this macro or the @code{eh_return} instruction pattern.
3418 @defmac RETURN_ADDR_OFFSET
3419 If defined, an integer-valued C expression for which rtl will be generated
3420 to add it to the exception handler address before it is searched in the
3421 exception handling tables, and to subtract it again from the address before
3422 using it to return to the exception handler.
3425 @defmac ASM_PREFERRED_EH_DATA_FORMAT (@var{code}, @var{global})
3426 This macro chooses the encoding of pointers embedded in the exception
3427 handling sections. If at all possible, this should be defined such
3428 that the exception handling section will not require dynamic relocations,
3429 and so may be read-only.
3431 @var{code} is 0 for data, 1 for code labels, 2 for function pointers.
3432 @var{global} is true if the symbol may be affected by dynamic relocations.
3433 The macro should return a combination of the @code{DW_EH_PE_*} defines
3434 as found in @file{dwarf2.h}.
3436 If this macro is not defined, pointers will not be encoded but
3437 represented directly.
3440 @defmac ASM_MAYBE_OUTPUT_ENCODED_ADDR_RTX (@var{file}, @var{encoding}, @var{size}, @var{addr}, @var{done})
3441 This macro allows the target to emit whatever special magic is required
3442 to represent the encoding chosen by @code{ASM_PREFERRED_EH_DATA_FORMAT}.
3443 Generic code takes care of pc-relative and indirect encodings; this must
3444 be defined if the target uses text-relative or data-relative encodings.
3446 This is a C statement that branches to @var{done} if the format was
3447 handled. @var{encoding} is the format chosen, @var{size} is the number
3448 of bytes that the format occupies, @var{addr} is the @code{SYMBOL_REF}
3452 @defmac MD_UNWIND_SUPPORT
3453 A string specifying a file to be #include'd in unwind-dw2.c. The file
3454 so included typically defines @code{MD_FALLBACK_FRAME_STATE_FOR}.
3457 @defmac MD_FALLBACK_FRAME_STATE_FOR (@var{context}, @var{fs})
3458 This macro allows the target to add CPU and operating system specific
3459 code to the call-frame unwinder for use when there is no unwind data
3460 available. The most common reason to implement this macro is to unwind
3461 through signal frames.
3463 This macro is called from @code{uw_frame_state_for} in
3464 @file{unwind-dw2.c}, @file{unwind-dw2-xtensa.c} and
3465 @file{unwind-ia64.c}. @var{context} is an @code{_Unwind_Context};
3466 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{context->ra}
3467 for the address of the code being executed and @code{context->cfa} for
3468 the stack pointer value. If the frame can be decoded, the register
3469 save addresses should be updated in @var{fs} and the macro should
3470 evaluate to @code{_URC_NO_REASON}. If the frame cannot be decoded,
3471 the macro should evaluate to @code{_URC_END_OF_STACK}.
3473 For proper signal handling in Java this macro is accompanied by
3474 @code{MAKE_THROW_FRAME}, defined in @file{libjava/include/*-signal.h} headers.
3477 @defmac MD_HANDLE_UNWABI (@var{context}, @var{fs})
3478 This macro allows the target to add operating system specific code to the
3479 call-frame unwinder to handle the IA-64 @code{.unwabi} unwinding directive,
3480 usually used for signal or interrupt frames.
3482 This macro is called from @code{uw_update_context} in @file{unwind-ia64.c}.
3483 @var{context} is an @code{_Unwind_Context};
3484 @var{fs} is an @code{_Unwind_FrameState}. Examine @code{fs->unwabi}
3485 for the abi and context in the @code{.unwabi} directive. If the
3486 @code{.unwabi} directive can be handled, the register save addresses should
3487 be updated in @var{fs}.
3490 @defmac TARGET_USES_WEAK_UNWIND_INFO
3491 A C expression that evaluates to true if the target requires unwind
3492 info to be given comdat linkage. Define it to be @code{1} if comdat
3493 linkage is necessary. The default is @code{0}.
3496 @node Stack Checking
3497 @subsection Specifying How Stack Checking is Done
3499 GCC will check that stack references are within the boundaries of the
3500 stack, if the option @option{-fstack-check} is specified, in one of
3505 If the value of the @code{STACK_CHECK_BUILTIN} macro is nonzero, GCC
3506 will assume that you have arranged for full stack checking to be done
3507 at appropriate places in the configuration files. GCC will not do
3508 other special processing.
3511 If @code{STACK_CHECK_BUILTIN} is zero and the value of the
3512 @code{STACK_CHECK_STATIC_BUILTIN} macro is nonzero, GCC will assume
3513 that you have arranged for static stack checking (checking of the
3514 static stack frame of functions) to be done at appropriate places
3515 in the configuration files. GCC will only emit code to do dynamic
3516 stack checking (checking on dynamic stack allocations) using the third
3520 If neither of the above are true, GCC will generate code to periodically
3521 ``probe'' the stack pointer using the values of the macros defined below.
3524 If neither STACK_CHECK_BUILTIN nor STACK_CHECK_STATIC_BUILTIN is defined,
3525 GCC will change its allocation strategy for large objects if the option
3526 @option{-fstack-check} is specified: they will always be allocated
3527 dynamically if their size exceeds @code{STACK_CHECK_MAX_VAR_SIZE} bytes.
3529 @defmac STACK_CHECK_BUILTIN
3530 A nonzero value if stack checking is done by the configuration files in a
3531 machine-dependent manner. You should define this macro if stack checking
3532 is require by the ABI of your machine or if you would like to do stack
3533 checking in some more efficient way than the generic approach. The default
3534 value of this macro is zero.
3537 @defmac STACK_CHECK_STATIC_BUILTIN
3538 A nonzero value if static stack checking is done by the configuration files
3539 in a machine-dependent manner. You should define this macro if you would
3540 like to do static stack checking in some more efficient way than the generic
3541 approach. The default value of this macro is zero.
3544 @defmac STACK_CHECK_PROBE_INTERVAL
3545 An integer representing the interval at which GCC must generate stack
3546 probe instructions. You will normally define this macro to be no larger
3547 than the size of the ``guard pages'' at the end of a stack area. The
3548 default value of 4096 is suitable for most systems.
3551 @defmac STACK_CHECK_PROBE_LOAD
3552 An integer which is nonzero if GCC should perform the stack probe
3553 as a load instruction and zero if GCC should use a store instruction.
3554 The default is zero, which is the most efficient choice on most systems.
3557 @defmac STACK_CHECK_PROTECT
3558 The number of bytes of stack needed to recover from a stack overflow,
3559 for languages where such a recovery is supported. The default value of
3560 75 words should be adequate for most machines.
3563 The following macros are relevant only if neither STACK_CHECK_BUILTIN
3564 nor STACK_CHECK_STATIC_BUILTIN is defined; you can omit them altogether
3565 in the opposite case.
3567 @defmac STACK_CHECK_MAX_FRAME_SIZE
3568 The maximum size of a stack frame, in bytes. GCC will generate probe
3569 instructions in non-leaf functions to ensure at least this many bytes of
3570 stack are available. If a stack frame is larger than this size, stack
3571 checking will not be reliable and GCC will issue a warning. The
3572 default is chosen so that GCC only generates one instruction on most
3573 systems. You should normally not change the default value of this macro.
3576 @defmac STACK_CHECK_FIXED_FRAME_SIZE
3577 GCC uses this value to generate the above warning message. It
3578 represents the amount of fixed frame used by a function, not including
3579 space for any callee-saved registers, temporaries and user variables.
3580 You need only specify an upper bound for this amount and will normally
3581 use the default of four words.
3584 @defmac STACK_CHECK_MAX_VAR_SIZE
3585 The maximum size, in bytes, of an object that GCC will place in the
3586 fixed area of the stack frame when the user specifies
3587 @option{-fstack-check}.
3588 GCC computed the default from the values of the above macros and you will
3589 normally not need to override that default.
3593 @node Frame Registers
3594 @subsection Registers That Address the Stack Frame
3596 @c prevent bad page break with this line
3597 This discusses registers that address the stack frame.
3599 @defmac STACK_POINTER_REGNUM
3600 The register number of the stack pointer register, which must also be a
3601 fixed register according to @code{FIXED_REGISTERS}. On most machines,
3602 the hardware determines which register this is.
3605 @defmac FRAME_POINTER_REGNUM
3606 The register number of the frame pointer register, which is used to
3607 access automatic variables in the stack frame. On some machines, the
3608 hardware determines which register this is. On other machines, you can
3609 choose any register you wish for this purpose.
3612 @defmac HARD_FRAME_POINTER_REGNUM
3613 On some machines the offset between the frame pointer and starting
3614 offset of the automatic variables is not known until after register
3615 allocation has been done (for example, because the saved registers are
3616 between these two locations). On those machines, define
3617 @code{FRAME_POINTER_REGNUM} the number of a special, fixed register to
3618 be used internally until the offset is known, and define
3619 @code{HARD_FRAME_POINTER_REGNUM} to be the actual hard register number
3620 used for the frame pointer.
3622 You should define this macro only in the very rare circumstances when it
3623 is not possible to calculate the offset between the frame pointer and
3624 the automatic variables until after register allocation has been
3625 completed. When this macro is defined, you must also indicate in your
3626 definition of @code{ELIMINABLE_REGS} how to eliminate
3627 @code{FRAME_POINTER_REGNUM} into either @code{HARD_FRAME_POINTER_REGNUM}
3628 or @code{STACK_POINTER_REGNUM}.
3630 Do not define this macro if it would be the same as
3631 @code{FRAME_POINTER_REGNUM}.
3634 @defmac ARG_POINTER_REGNUM
3635 The register number of the arg pointer register, which is used to access
3636 the function's argument list. On some machines, this is the same as the
3637 frame pointer register. On some machines, the hardware determines which
3638 register this is. On other machines, you can choose any register you
3639 wish for this purpose. If this is not the same register as the frame
3640 pointer register, then you must mark it as a fixed register according to
3641 @code{FIXED_REGISTERS}, or arrange to be able to eliminate it
3642 (@pxref{Elimination}).
3645 @defmac RETURN_ADDRESS_POINTER_REGNUM
3646 The register number of the return address pointer register, which is used to
3647 access the current function's return address from the stack. On some
3648 machines, the return address is not at a fixed offset from the frame
3649 pointer or stack pointer or argument pointer. This register can be defined
3650 to point to the return address on the stack, and then be converted by
3651 @code{ELIMINABLE_REGS} into either the frame pointer or stack pointer.
3653 Do not define this macro unless there is no other way to get the return
3654 address from the stack.
3657 @defmac STATIC_CHAIN_REGNUM
3658 @defmacx STATIC_CHAIN_INCOMING_REGNUM
3659 Register numbers used for passing a function's static chain pointer. If
3660 register windows are used, the register number as seen by the called
3661 function is @code{STATIC_CHAIN_INCOMING_REGNUM}, while the register
3662 number as seen by the calling function is @code{STATIC_CHAIN_REGNUM}. If
3663 these registers are the same, @code{STATIC_CHAIN_INCOMING_REGNUM} need
3666 The static chain register need not be a fixed register.
3668 If the static chain is passed in memory, these macros should not be
3669 defined; instead, the @code{TARGET_STATIC_CHAIN} hook should be used.
3672 @deftypefn {Target Hook} rtx TARGET_STATIC_CHAIN (const_tree @var{fndecl}, bool @var{incoming_p})
3673 This hook replaces the use of @code{STATIC_CHAIN_REGNUM} et al for
3674 targets that may use different static chain locations for different
3675 nested functions. This may be required if the target has function
3676 attributes that affect the calling conventions of the function and
3677 those calling conventions use different static chain locations.
3679 The default version of this hook uses @code{STATIC_CHAIN_REGNUM} et al.
3681 If the static chain is passed in memory, this hook should be used to
3682 provide rtx giving @code{mem} expressions that denote where they are stored.
3683 Often the @code{mem} expression as seen by the caller will be at an offset
3684 from the stack pointer and the @code{mem} expression as seen by the callee
3685 will be at an offset from the frame pointer.
3686 @findex stack_pointer_rtx
3687 @findex frame_pointer_rtx
3688 @findex arg_pointer_rtx
3689 The variables @code{stack_pointer_rtx}, @code{frame_pointer_rtx}, and
3690 @code{arg_pointer_rtx} will have been initialized and should be used
3691 to refer to those items.
3694 @defmac DWARF_FRAME_REGISTERS
3695 This macro specifies the maximum number of hard registers that can be
3696 saved in a call frame. This is used to size data structures used in
3697 DWARF2 exception handling.
3699 Prior to GCC 3.0, this macro was needed in order to establish a stable
3700 exception handling ABI in the face of adding new hard registers for ISA
3701 extensions. In GCC 3.0 and later, the EH ABI is insulated from changes
3702 in the number of hard registers. Nevertheless, this macro can still be
3703 used to reduce the runtime memory requirements of the exception handling
3704 routines, which can be substantial if the ISA contains a lot of
3705 registers that are not call-saved.
3707 If this macro is not defined, it defaults to
3708 @code{FIRST_PSEUDO_REGISTER}.
3711 @defmac PRE_GCC3_DWARF_FRAME_REGISTERS
3713 This macro is similar to @code{DWARF_FRAME_REGISTERS}, but is provided
3714 for backward compatibility in pre GCC 3.0 compiled code.
3716 If this macro is not defined, it defaults to
3717 @code{DWARF_FRAME_REGISTERS}.
3720 @defmac DWARF_REG_TO_UNWIND_COLUMN (@var{regno})
3722 Define this macro if the target's representation for dwarf registers
3723 is different than the internal representation for unwind column.
3724 Given a dwarf register, this macro should return the internal unwind
3725 column number to use instead.
3727 See the PowerPC's SPE target for an example.
3730 @defmac DWARF_FRAME_REGNUM (@var{regno})
3732 Define this macro if the target's representation for dwarf registers
3733 used in .eh_frame or .debug_frame is different from that used in other
3734 debug info sections. Given a GCC hard register number, this macro
3735 should return the .eh_frame register number. The default is
3736 @code{DBX_REGISTER_NUMBER (@var{regno})}.
3740 @defmac DWARF2_FRAME_REG_OUT (@var{regno}, @var{for_eh})
3742 Define this macro to map register numbers held in the call frame info
3743 that GCC has collected using @code{DWARF_FRAME_REGNUM} to those that
3744 should be output in .debug_frame (@code{@var{for_eh}} is zero) and
3745 .eh_frame (@code{@var{for_eh}} is nonzero). The default is to
3746 return @code{@var{regno}}.
3751 @subsection Eliminating Frame Pointer and Arg Pointer
3753 @c prevent bad page break with this line
3754 This is about eliminating the frame pointer and arg pointer.
3756 @deftypefn {Target Hook} bool TARGET_FRAME_POINTER_REQUIRED (void)
3757 This target hook should return @code{true} if a function must have and use
3758 a frame pointer. This target hook is called in the reload pass. If its return
3759 value is @code{true} the function will have a frame pointer.
3761 This target hook can in principle examine the current function and decide
3762 according to the facts, but on most machines the constant @code{false} or the
3763 constant @code{true} suffices. Use @code{false} when the machine allows code
3764 to be generated with no frame pointer, and doing so saves some time or space.
3765 Use @code{true} when there is no possible advantage to avoiding a frame
3768 In certain cases, the compiler does not know how to produce valid code
3769 without a frame pointer. The compiler recognizes those cases and
3770 automatically gives the function a frame pointer regardless of what
3771 @code{TARGET_FRAME_POINTER_REQUIRED} returns. You don't need to worry about
3774 In a function that does not require a frame pointer, the frame pointer
3775 register can be allocated for ordinary usage, unless you mark it as a
3776 fixed register. See @code{FIXED_REGISTERS} for more information.
3778 Default return value is @code{false}.
3781 @findex get_frame_size
3782 @defmac INITIAL_FRAME_POINTER_OFFSET (@var{depth-var})
3783 A C statement to store in the variable @var{depth-var} the difference
3784 between the frame pointer and the stack pointer values immediately after
3785 the function prologue. The value would be computed from information
3786 such as the result of @code{get_frame_size ()} and the tables of
3787 registers @code{regs_ever_live} and @code{call_used_regs}.
3789 If @code{ELIMINABLE_REGS} is defined, this macro will be not be used and
3790 need not be defined. Otherwise, it must be defined even if
3791 @code{TARGET_FRAME_POINTER_REQUIRED} is always return true; in that
3792 case, you may set @var{depth-var} to anything.
3795 @defmac ELIMINABLE_REGS
3796 If defined, this macro specifies a table of register pairs used to
3797 eliminate unneeded registers that point into the stack frame. If it is not
3798 defined, the only elimination attempted by the compiler is to replace
3799 references to the frame pointer with references to the stack pointer.
3801 The definition of this macro is a list of structure initializations, each
3802 of which specifies an original and replacement register.
3804 On some machines, the position of the argument pointer is not known until
3805 the compilation is completed. In such a case, a separate hard register
3806 must be used for the argument pointer. This register can be eliminated by
3807 replacing it with either the frame pointer or the argument pointer,
3808 depending on whether or not the frame pointer has been eliminated.
3810 In this case, you might specify:
3812 #define ELIMINABLE_REGS \
3813 @{@{ARG_POINTER_REGNUM, STACK_POINTER_REGNUM@}, \
3814 @{ARG_POINTER_REGNUM, FRAME_POINTER_REGNUM@}, \
3815 @{FRAME_POINTER_REGNUM, STACK_POINTER_REGNUM@}@}
3818 Note that the elimination of the argument pointer with the stack pointer is
3819 specified first since that is the preferred elimination.
3822 @deftypefn {Target Hook} bool TARGET_CAN_ELIMINATE (const int @var{from-reg}, const int @var{to-reg})
3823 This target hook should returns @code{true} if the compiler is allowed to
3824 try to replace register number @var{from-reg} with register number
3825 @var{to-reg}. This target hook need only be defined if @code{ELIMINABLE_REGS}
3826 is defined, and will usually be @code{true}, since most of the cases
3827 preventing register elimination are things that the compiler already
3830 Default return value is @code{true}.
3833 @defmac INITIAL_ELIMINATION_OFFSET (@var{from-reg}, @var{to-reg}, @var{offset-var})
3834 This macro is similar to @code{INITIAL_FRAME_POINTER_OFFSET}. It
3835 specifies the initial difference between the specified pair of
3836 registers. This macro must be defined if @code{ELIMINABLE_REGS} is
3840 @node Stack Arguments
3841 @subsection Passing Function Arguments on the Stack
3842 @cindex arguments on stack
3843 @cindex stack arguments
3845 The macros in this section control how arguments are passed
3846 on the stack. See the following section for other macros that
3847 control passing certain arguments in registers.
3849 @deftypefn {Target Hook} bool TARGET_PROMOTE_PROTOTYPES (tree @var{fntype})
3850 This target hook returns @code{true} if an argument declared in a
3851 prototype as an integral type smaller than @code{int} should actually be
3852 passed as an @code{int}. In addition to avoiding errors in certain
3853 cases of mismatch, it also makes for better code on certain machines.
3854 The default is to not promote prototypes.
3858 A C expression. If nonzero, push insns will be used to pass
3860 If the target machine does not have a push instruction, set it to zero.
3861 That directs GCC to use an alternate strategy: to
3862 allocate the entire argument block and then store the arguments into
3863 it. When @code{PUSH_ARGS} is nonzero, @code{PUSH_ROUNDING} must be defined too.
3866 @defmac PUSH_ARGS_REVERSED
3867 A C expression. If nonzero, function arguments will be evaluated from
3868 last to first, rather than from first to last. If this macro is not
3869 defined, it defaults to @code{PUSH_ARGS} on targets where the stack
3870 and args grow in opposite directions, and 0 otherwise.
3873 @defmac PUSH_ROUNDING (@var{npushed})
3874 A C expression that is the number of bytes actually pushed onto the
3875 stack when an instruction attempts to push @var{npushed} bytes.
3877 On some machines, the definition
3880 #define PUSH_ROUNDING(BYTES) (BYTES)
3884 will suffice. But on other machines, instructions that appear
3885 to push one byte actually push two bytes in an attempt to maintain
3886 alignment. Then the definition should be
3889 #define PUSH_ROUNDING(BYTES) (((BYTES) + 1) & ~1)
3893 @findex current_function_outgoing_args_size
3894 @defmac ACCUMULATE_OUTGOING_ARGS
3895 A C expression. If nonzero, the maximum amount of space required for outgoing arguments
3896 will be computed and placed into the variable
3897 @code{current_function_outgoing_args_size}. No space will be pushed
3898 onto the stack for each call; instead, the function prologue should
3899 increase the stack frame size by this amount.
3901 Setting both @code{PUSH_ARGS} and @code{ACCUMULATE_OUTGOING_ARGS}
3905 @defmac REG_PARM_STACK_SPACE (@var{fndecl})
3906 Define this macro if functions should assume that stack space has been
3907 allocated for arguments even when their values are passed in
3910 The value of this macro is the size, in bytes, of the area reserved for
3911 arguments passed in registers for the function represented by @var{fndecl},
3912 which can be zero if GCC is calling a library function.
3913 The argument @var{fndecl} can be the FUNCTION_DECL, or the type itself
3916 This space can be allocated by the caller, or be a part of the
3917 machine-dependent stack frame: @code{OUTGOING_REG_PARM_STACK_SPACE} says
3920 @c above is overfull. not sure what to do. --mew 5feb93 did
3921 @c something, not sure if it looks good. --mew 10feb93
3923 @defmac OUTGOING_REG_PARM_STACK_SPACE (@var{fntype})
3924 Define this to a nonzero value if it is the responsibility of the
3925 caller to allocate the area reserved for arguments passed in registers
3926 when calling a function of @var{fntype}. @var{fntype} may be NULL
3927 if the function called is a library function.
3929 If @code{ACCUMULATE_OUTGOING_ARGS} is defined, this macro controls
3930 whether the space for these arguments counts in the value of
3931 @code{current_function_outgoing_args_size}.
3934 @defmac STACK_PARMS_IN_REG_PARM_AREA
3935 Define this macro if @code{REG_PARM_STACK_SPACE} is defined, but the
3936 stack parameters don't skip the area specified by it.
3937 @c i changed this, makes more sens and it should have taken care of the
3938 @c overfull.. not as specific, tho. --mew 5feb93
3940 Normally, when a parameter is not passed in registers, it is placed on the
3941 stack beyond the @code{REG_PARM_STACK_SPACE} area. Defining this macro
3942 suppresses this behavior and causes the parameter to be passed on the
3943 stack in its natural location.
3946 @defmac RETURN_POPS_ARGS (@var{fundecl}, @var{funtype}, @var{stack-size})
3947 A C expression that should indicate the number of bytes of its own
3948 arguments that a function pops on returning, or 0 if the
3949 function pops no arguments and the caller must therefore pop them all
3950 after the function returns.
3952 @var{fundecl} is a C variable whose value is a tree node that describes
3953 the function in question. Normally it is a node of type
3954 @code{FUNCTION_DECL} that describes the declaration of the function.
3955 From this you can obtain the @code{DECL_ATTRIBUTES} of the function.
3957 @var{funtype} is a C variable whose value is a tree node that
3958 describes the function in question. Normally it is a node of type
3959 @code{FUNCTION_TYPE} that describes the data type of the function.
3960 From this it is possible to obtain the data types of the value and
3961 arguments (if known).
3963 When a call to a library function is being considered, @var{fundecl}
3964 will contain an identifier node for the library function. Thus, if
3965 you need to distinguish among various library functions, you can do so
3966 by their names. Note that ``library function'' in this context means
3967 a function used to perform arithmetic, whose name is known specially
3968 in the compiler and was not mentioned in the C code being compiled.
3970 @var{stack-size} is the number of bytes of arguments passed on the
3971 stack. If a variable number of bytes is passed, it is zero, and
3972 argument popping will always be the responsibility of the calling function.
3974 On the VAX, all functions always pop their arguments, so the definition
3975 of this macro is @var{stack-size}. On the 68000, using the standard
3976 calling convention, no functions pop their arguments, so the value of
3977 the macro is always 0 in this case. But an alternative calling
3978 convention is available in which functions that take a fixed number of
3979 arguments pop them but other functions (such as @code{printf}) pop
3980 nothing (the caller pops all). When this convention is in use,
3981 @var{funtype} is examined to determine whether a function takes a fixed
3982 number of arguments.
3985 @defmac CALL_POPS_ARGS (@var{cum})
3986 A C expression that should indicate the number of bytes a call sequence
3987 pops off the stack. It is added to the value of @code{RETURN_POPS_ARGS}
3988 when compiling a function call.
3990 @var{cum} is the variable in which all arguments to the called function
3991 have been accumulated.
3993 On certain architectures, such as the SH5, a call trampoline is used
3994 that pops certain registers off the stack, depending on the arguments
3995 that have been passed to the function. Since this is a property of the
3996 call site, not of the called function, @code{RETURN_POPS_ARGS} is not
4000 @node Register Arguments
4001 @subsection Passing Arguments in Registers
4002 @cindex arguments in registers
4003 @cindex registers arguments
4005 This section describes the macros which let you control how various
4006 types of arguments are passed in registers or how they are arranged in
4009 @defmac FUNCTION_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
4010 A C expression that controls whether a function argument is passed
4011 in a register, and which register.
4013 The arguments are @var{cum}, which summarizes all the previous
4014 arguments; @var{mode}, the machine mode of the argument; @var{type},
4015 the data type of the argument as a tree node or 0 if that is not known
4016 (which happens for C support library functions); and @var{named},
4017 which is 1 for an ordinary argument and 0 for nameless arguments that
4018 correspond to @samp{@dots{}} in the called function's prototype.
4019 @var{type} can be an incomplete type if a syntax error has previously
4022 The value of the expression is usually either a @code{reg} RTX for the
4023 hard register in which to pass the argument, or zero to pass the
4024 argument on the stack.
4026 For machines like the VAX and 68000, where normally all arguments are
4027 pushed, zero suffices as a definition.
4029 The value of the expression can also be a @code{parallel} RTX@. This is
4030 used when an argument is passed in multiple locations. The mode of the
4031 @code{parallel} should be the mode of the entire argument. The
4032 @code{parallel} holds any number of @code{expr_list} pairs; each one
4033 describes where part of the argument is passed. In each
4034 @code{expr_list} the first operand must be a @code{reg} RTX for the hard
4035 register in which to pass this part of the argument, and the mode of the
4036 register RTX indicates how large this part of the argument is. The
4037 second operand of the @code{expr_list} is a @code{const_int} which gives
4038 the offset in bytes into the entire argument of where this part starts.
4039 As a special exception the first @code{expr_list} in the @code{parallel}
4040 RTX may have a first operand of zero. This indicates that the entire
4041 argument is also stored on the stack.
4043 The last time this macro is called, it is called with @code{MODE ==
4044 VOIDmode}, and its result is passed to the @code{call} or @code{call_value}
4045 pattern as operands 2 and 3 respectively.
4047 @cindex @file{stdarg.h} and register arguments
4048 The usual way to make the ISO library @file{stdarg.h} work on a machine
4049 where some arguments are usually passed in registers, is to cause
4050 nameless arguments to be passed on the stack instead. This is done
4051 by making @code{FUNCTION_ARG} return 0 whenever @var{named} is 0.
4053 @cindex @code{TARGET_MUST_PASS_IN_STACK}, and @code{FUNCTION_ARG}
4054 @cindex @code{REG_PARM_STACK_SPACE}, and @code{FUNCTION_ARG}
4055 You may use the hook @code{targetm.calls.must_pass_in_stack}
4056 in the definition of this macro to determine if this argument is of a
4057 type that must be passed in the stack. If @code{REG_PARM_STACK_SPACE}
4058 is not defined and @code{FUNCTION_ARG} returns nonzero for such an
4059 argument, the compiler will abort. If @code{REG_PARM_STACK_SPACE} is
4060 defined, the argument will be computed in the stack and then loaded into
4064 @deftypefn {Target Hook} bool TARGET_MUST_PASS_IN_STACK (enum machine_mode @var{mode}, tree @var{type})
4065 This target hook should return @code{true} if we should not pass @var{type}
4066 solely in registers. The file @file{expr.h} defines a
4067 definition that is usually appropriate, refer to @file{expr.h} for additional
4071 @defmac FUNCTION_INCOMING_ARG (@var{cum}, @var{mode}, @var{type}, @var{named})
4072 Define this macro if the target machine has ``register windows'', so
4073 that the register in which a function sees an arguments is not
4074 necessarily the same as the one in which the caller passed the
4077 For such machines, @code{FUNCTION_ARG} computes the register in which
4078 the caller passes the value, and @code{FUNCTION_INCOMING_ARG} should
4079 be defined in a similar fashion to tell the function being called
4080 where the arguments will arrive.
4082 If @code{FUNCTION_INCOMING_ARG} is not defined, @code{FUNCTION_ARG}
4083 serves both purposes.
4086 @deftypefn {Target Hook} int TARGET_ARG_PARTIAL_BYTES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4087 This target hook returns the number of bytes at the beginning of an
4088 argument that must be put in registers. The value must be zero for
4089 arguments that are passed entirely in registers or that are entirely
4090 pushed on the stack.
4092 On some machines, certain arguments must be passed partially in
4093 registers and partially in memory. On these machines, typically the
4094 first few words of arguments are passed in registers, and the rest
4095 on the stack. If a multi-word argument (a @code{double} or a
4096 structure) crosses that boundary, its first few words must be passed
4097 in registers and the rest must be pushed. This macro tells the
4098 compiler when this occurs, and how many bytes should go in registers.
4100 @code{FUNCTION_ARG} for these arguments should return the first
4101 register to be used by the caller for this argument; likewise
4102 @code{FUNCTION_INCOMING_ARG}, for the called function.
4105 @deftypefn {Target Hook} bool TARGET_PASS_BY_REFERENCE (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4106 This target hook should return @code{true} if an argument at the
4107 position indicated by @var{cum} should be passed by reference. This
4108 predicate is queried after target independent reasons for being
4109 passed by reference, such as @code{TREE_ADDRESSABLE (type)}.
4111 If the hook returns true, a copy of that argument is made in memory and a
4112 pointer to the argument is passed instead of the argument itself.
4113 The pointer is passed in whatever way is appropriate for passing a pointer
4117 @deftypefn {Target Hook} bool TARGET_CALLEE_COPIES (CUMULATIVE_ARGS *@var{cum}, enum machine_mode @var{mode}, tree @var{type}, bool @var{named})
4118 The function argument described by the parameters to this hook is
4119 known to be passed by reference. The hook should return true if the
4120 function argument should be copied by the callee instead of copied
4123 For any argument for which the hook returns true, if it can be
4124 determined that the argument is not modified, then a copy need
4127 The default version of this hook always returns false.
4130 @defmac CUMULATIVE_ARGS
4131 A C type for declaring a variable that is used as the first argument of
4132 @code{FUNCTION_ARG} and other related values. For some target machines,
4133 the type @code{int} suffices and can hold the number of bytes of
4136 There is no need to record in @code{CUMULATIVE_ARGS} anything about the
4137 arguments that have been passed on the stack. The compiler has other
4138 variables to keep track of that. For target machines on which all
4139 arguments are passed on the stack, there is no need to store anything in
4140 @code{CUMULATIVE_ARGS}; however, the data structure must exist and
4141 should not be empty, so use @code{int}.
4144 @defmac OVERRIDE_ABI_FORMAT (@var{fndecl})
4145 If defined, this macro is called before generating any code for a
4146 function, but after the @var{cfun} descriptor for the function has been
4147 created. The back end may use this macro to update @var{cfun} to
4148 reflect an ABI other than that which would normally be used by default.
4149 If the compiler is generating code for a compiler-generated function,
4150 @var{fndecl} may be @code{NULL}.
4153 @defmac INIT_CUMULATIVE_ARGS (@var{cum}, @var{fntype}, @var{libname}, @var{fndecl}, @var{n_named_args})
4154 A C statement (sans semicolon) for initializing the variable
4155 @var{cum} for the state at the beginning of the argument list. The
4156 variable has type @code{CUMULATIVE_ARGS}. The value of @var{fntype}
4157 is the tree node for the data type of the function which will receive
4158 the args, or 0 if the args are to a compiler support library function.
4159 For direct calls that are not libcalls, @var{fndecl} contain the
4160 declaration node of the function. @var{fndecl} is also set when
4161 @code{INIT_CUMULATIVE_ARGS} is used to find arguments for the function
4162 being compiled. @var{n_named_args} is set to the number of named
4163 arguments, including a structure return address if it is passed as a
4164 parameter, when making a call. When processing incoming arguments,
4165 @var{n_named_args} is set to @minus{}1.
4167 When processing a call to a compiler support library function,
4168 @var{libname} identifies which one. It is a @code{symbol_ref} rtx which
4169 contains the name of the function, as a string. @var{libname} is 0 when
4170 an ordinary C function call is being processed. Thus, each time this
4171 macro is called, either @var{libname} or @var{fntype} is nonzero, but
4172 never both of them at once.
4175 @defmac INIT_CUMULATIVE_LIBCALL_ARGS (@var{cum}, @var{mode}, @var{libname})
4176 Like @code{INIT_CUMULATIVE_ARGS} but only used for outgoing libcalls,
4177 it gets a @code{MODE} argument instead of @var{fntype}, that would be
4178 @code{NULL}. @var{indirect} would always be zero, too. If this macro
4179 is not defined, @code{INIT_CUMULATIVE_ARGS (cum, NULL_RTX, libname,
4180 0)} is used instead.
4183 @defmac INIT_CUMULATIVE_INCOMING_ARGS (@var{cum}, @var{fntype}, @var{libname})
4184 Like @code{INIT_CUMULATIVE_ARGS} but overrides it for the purposes of
4185 finding the arguments for the function being compiled. If this macro is
4186 undefined, @code{INIT_CUMULATIVE_ARGS} is used instead.
4188 The value passed for @var{libname} is always 0, since library routines
4189 with special calling conventions are never compiled with GCC@. The
4190 argument @var{libname} exists for symmetry with
4191 @code{INIT_CUMULATIVE_ARGS}.
4192 @c could use "this macro" in place of @code{INIT_CUMULATIVE_ARGS}, maybe.
4193 @c --mew 5feb93 i switched the order of the sentences. --mew 10feb93
4196 @defmac FUNCTION_ARG_ADVANCE (@var{cum}, @var{mode}, @var{type}, @var{named})
4197 A C statement (sans semicolon) to update the summarizer variable
4198 @var{cum} to advance past an argument in the argument list. The
4199 values @var{mode}, @var{type} and @var{named} describe that argument.
4200 Once this is done, the variable @var{cum} is suitable for analyzing
4201 the @emph{following} argument with @code{FUNCTION_ARG}, etc.
4203 This macro need not do anything if the argument in question was passed
4204 on the stack. The compiler knows how to track the amount of stack space
4205 used for arguments without any special help.
4209 @defmac FUNCTION_ARG_OFFSET (@var{mode}, @var{type})
4210 If defined, a C expression that is the number of bytes to add to the
4211 offset of the argument passed in memory. This is needed for the SPU,
4212 which passes @code{char} and @code{short} arguments in the preferred
4213 slot that is in the middle of the quad word instead of starting at the
4217 @defmac FUNCTION_ARG_PADDING (@var{mode}, @var{type})
4218 If defined, a C expression which determines whether, and in which direction,
4219 to pad out an argument with extra space. The value should be of type
4220 @code{enum direction}: either @code{upward} to pad above the argument,
4221 @code{downward} to pad below, or @code{none} to inhibit padding.
4223 The @emph{amount} of padding is always just enough to reach the next
4224 multiple of @code{FUNCTION_ARG_BOUNDARY}; this macro does not control
4227 This macro has a default definition which is right for most systems.
4228 For little-endian machines, the default is to pad upward. For
4229 big-endian machines, the default is to pad downward for an argument of
4230 constant size shorter than an @code{int}, and upward otherwise.
4233 @defmac PAD_VARARGS_DOWN
4234 If defined, a C expression which determines whether the default
4235 implementation of va_arg will attempt to pad down before reading the
4236 next argument, if that argument is smaller than its aligned space as
4237 controlled by @code{PARM_BOUNDARY}. If this macro is not defined, all such
4238 arguments are padded down if @code{BYTES_BIG_ENDIAN} is true.
4241 @defmac BLOCK_REG_PADDING (@var{mode}, @var{type}, @var{first})
4242 Specify padding for the last element of a block move between registers and
4243 memory. @var{first} is nonzero if this is the only element. Defining this
4244 macro allows better control of register function parameters on big-endian
4245 machines, without using @code{PARALLEL} rtl. In particular,
4246 @code{MUST_PASS_IN_STACK} need not test padding and mode of types in
4247 registers, as there is no longer a "wrong" part of a register; For example,
4248 a three byte aggregate may be passed in the high part of a register if so
4252 @defmac FUNCTION_ARG_BOUNDARY (@var{mode}, @var{type})
4253 If defined, a C expression that gives the alignment boundary, in bits,
4254 of an argument with the specified mode and type. If it is not defined,
4255 @code{PARM_BOUNDARY} is used for all arguments.
4258 @defmac FUNCTION_ARG_REGNO_P (@var{regno})
4259 A C expression that is nonzero if @var{regno} is the number of a hard
4260 register in which function arguments are sometimes passed. This does
4261 @emph{not} include implicit arguments such as the static chain and
4262 the structure-value address. On many machines, no registers can be
4263 used for this purpose since all function arguments are pushed on the
4267 @deftypefn {Target Hook} bool TARGET_SPLIT_COMPLEX_ARG (tree @var{type})
4268 This hook should return true if parameter of type @var{type} are passed
4269 as two scalar parameters. By default, GCC will attempt to pack complex
4270 arguments into the target's word size. Some ABIs require complex arguments
4271 to be split and treated as their individual components. For example, on
4272 AIX64, complex floats should be passed in a pair of floating point
4273 registers, even though a complex float would fit in one 64-bit floating
4276 The default value of this hook is @code{NULL}, which is treated as always
4280 @deftypefn {Target Hook} tree TARGET_BUILD_BUILTIN_VA_LIST (void)
4281 This hook returns a type node for @code{va_list} for the target.
4282 The default version of the hook returns @code{void*}.
4285 @deftypefn {Target Hook} tree TARGET_FN_ABI_VA_LIST (tree @var{fndecl})
4286 This hook returns the va_list type of the calling convention specified by
4288 The default version of this hook returns @code{va_list_type_node}.
4291 @deftypefn {Target Hook} tree TARGET_CANONICAL_VA_LIST_TYPE (tree @var{type})
4292 This hook returns the va_list type of the calling convention specified by the
4293 type of @var{type}. If @var{type} is not a valid va_list type, it returns
4297 @deftypefn {Target Hook} tree TARGET_GIMPLIFY_VA_ARG_EXPR (tree @var{valist}, tree @var{type}, tree *@var{pre_p}, tree *@var{post_p})
4298 This hook performs target-specific gimplification of
4299 @code{VA_ARG_EXPR}. The first two parameters correspond to the
4300 arguments to @code{va_arg}; the latter two are as in
4301 @code{gimplify.c:gimplify_expr}.
4304 @deftypefn {Target Hook} bool TARGET_VALID_POINTER_MODE (enum machine_mode @var{mode})
4305 Define this to return nonzero if the port can handle pointers
4306 with machine mode @var{mode}. The default version of this
4307 hook returns true for both @code{ptr_mode} and @code{Pmode}.
4310 @deftypefn {Target Hook} bool TARGET_SCALAR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4311 Define this to return nonzero if the port is prepared to handle
4312 insns involving scalar mode @var{mode}. For a scalar mode to be
4313 considered supported, all the basic arithmetic and comparisons
4316 The default version of this hook returns true for any mode
4317 required to handle the basic C types (as defined by the port).
4318 Included here are the double-word arithmetic supported by the
4319 code in @file{optabs.c}.
4322 @deftypefn {Target Hook} bool TARGET_VECTOR_MODE_SUPPORTED_P (enum machine_mode @var{mode})
4323 Define this to return nonzero if the port is prepared to handle
4324 insns involving vector mode @var{mode}. At the very least, it
4325 must have move patterns for this mode.
4329 @subsection How Scalar Function Values Are Returned
4330 @cindex return values in registers
4331 @cindex values, returned by functions
4332 @cindex scalars, returned as values
4334 This section discusses the macros that control returning scalars as
4335 values---values that can fit in registers.
4337 @deftypefn {Target Hook} rtx TARGET_FUNCTION_VALUE (tree @var{ret_type}, tree @var{fn_decl_or_type}, bool @var{outgoing})
4339 Define this to return an RTX representing the place where a function
4340 returns or receives a value of data type @var{ret_type}, a tree node
4341 representing a data type. @var{fn_decl_or_type} is a tree node
4342 representing @code{FUNCTION_DECL} or @code{FUNCTION_TYPE} of a
4343 function being called. If @var{outgoing} is false, the hook should
4344 compute the register in which the caller will see the return value.
4345 Otherwise, the hook should return an RTX representing the place where
4346 a function returns a value.
4348 On many machines, only @code{TYPE_MODE (@var{ret_type})} is relevant.
4349 (Actually, on most machines, scalar values are returned in the same
4350 place regardless of mode.) The value of the expression is usually a
4351 @code{reg} RTX for the hard register where the return value is stored.
4352 The value can also be a @code{parallel} RTX, if the return value is in
4353 multiple places. See @code{FUNCTION_ARG} for an explanation of the
4354 @code{parallel} form. Note that the callee will populate every
4355 location specified in the @code{parallel}, but if the first element of
4356 the @code{parallel} contains the whole return value, callers will use
4357 that element as the canonical location and ignore the others. The m68k
4358 port uses this type of @code{parallel} to return pointers in both
4359 @samp{%a0} (the canonical location) and @samp{%d0}.
4361 If @code{TARGET_PROMOTE_FUNCTION_RETURN} returns true, you must apply
4362 the same promotion rules specified in @code{PROMOTE_MODE} if
4363 @var{valtype} is a scalar type.
4365 If the precise function being called is known, @var{func} is a tree
4366 node (@code{FUNCTION_DECL}) for it; otherwise, @var{func} is a null
4367 pointer. This makes it possible to use a different value-returning
4368 convention for specific functions when all their calls are
4371 Some target machines have ``register windows'' so that the register in
4372 which a function returns its value is not the same as the one in which
4373 the caller sees the value. For such machines, you should return
4374 different RTX depending on @var{outgoing}.
4376 @code{TARGET_FUNCTION_VALUE} is not used for return values with
4377 aggregate data types, because these are returned in another way. See
4378 @code{TARGET_STRUCT_VALUE_RTX} and related macros, below.
4381 @defmac FUNCTION_VALUE (@var{valtype}, @var{func})
4382 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4383 a new target instead.
4386 @defmac FUNCTION_OUTGOING_VALUE (@var{valtype}, @var{func})
4387 This macro has been deprecated. Use @code{TARGET_FUNCTION_VALUE} for
4388 a new target instead.
4391 @defmac LIBCALL_VALUE (@var{mode})
4392 A C expression to create an RTX representing the place where a library
4393 function returns a value of mode @var{mode}.
4395 Note that ``library function'' in this context means a compiler
4396 support routine, used to perform arithmetic, whose name is known
4397 specially by the compiler and was not mentioned in the C code being
4401 @deftypefn {Target Hook} rtx TARGET_LIBCALL_VALUE (enum machine_mode
4402 @var{mode}, const_rtx @var{fun})
4403 Define this hook if the back-end needs to know the name of the libcall
4404 function in order to determine where the result should be returned.
4406 The mode of the result is given by @var{mode} and the name of the called
4407 library function is given by @var{fun}. The hook should return an RTX
4408 representing the place where the library function result will be returned.
4410 If this hook is not defined, then LIBCALL_VALUE will be used.
4413 @defmac FUNCTION_VALUE_REGNO_P (@var{regno})
4414 A C expression that is nonzero if @var{regno} is the number of a hard
4415 register in which the values of called function may come back.
4417 A register whose use for returning values is limited to serving as the
4418 second of a pair (for a value of type @code{double}, say) need not be
4419 recognized by this macro. So for most machines, this definition
4423 #define FUNCTION_VALUE_REGNO_P(N) ((N) == 0)
4426 If the machine has register windows, so that the caller and the called
4427 function use different registers for the return value, this macro
4428 should recognize only the caller's register numbers.
4431 @defmac TARGET_ENUM_VA_LIST (@var{idx}, @var{pname}, @var{ptype})
4432 This target macro is used in function @code{c_common_nodes_and_builtins}
4433 to iterate through the target specific builtin types for va_list. The
4434 variable @var{idx} is used as iterator. @var{pname} has to be a pointer
4435 to a @code{const char *} and @var{ptype} a pointer to a @code{tree} typed
4437 The arguments @var{pname} and @var{ptype} are used to store the result of
4438 this macro and are set to the name of the va_list builtin type and its
4440 If the return value of this macro is zero, then there is no more element.
4441 Otherwise the @var{IDX} should be increased for the next call of this
4442 macro to iterate through all types.
4445 @defmac APPLY_RESULT_SIZE
4446 Define this macro if @samp{untyped_call} and @samp{untyped_return}
4447 need more space than is implied by @code{FUNCTION_VALUE_REGNO_P} for
4448 saving and restoring an arbitrary return value.
4451 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MSB (tree @var{type})
4452 This hook should return true if values of type @var{type} are returned
4453 at the most significant end of a register (in other words, if they are
4454 padded at the least significant end). You can assume that @var{type}
4455 is returned in a register; the caller is required to check this.
4457 Note that the register provided by @code{TARGET_FUNCTION_VALUE} must
4458 be able to hold the complete return value. For example, if a 1-, 2-
4459 or 3-byte structure is returned at the most significant end of a
4460 4-byte register, @code{TARGET_FUNCTION_VALUE} should provide an
4464 @node Aggregate Return
4465 @subsection How Large Values Are Returned
4466 @cindex aggregates as return values
4467 @cindex large return values
4468 @cindex returning aggregate values
4469 @cindex structure value address
4471 When a function value's mode is @code{BLKmode} (and in some other
4472 cases), the value is not returned according to
4473 @code{TARGET_FUNCTION_VALUE} (@pxref{Scalar Return}). Instead, the
4474 caller passes the address of a block of memory in which the value
4475 should be stored. This address is called the @dfn{structure value
4478 This section describes how to control returning structure values in
4481 @deftypefn {Target Hook} bool TARGET_RETURN_IN_MEMORY (tree @var{type}, tree @var{fntype})
4482 This target hook should return a nonzero value to say to return the
4483 function value in memory, just as large structures are always returned.
4484 Here @var{type} will be the data type of the value, and @var{fntype}
4485 will be the type of the function doing the returning, or @code{NULL} for
4488 Note that values of mode @code{BLKmode} must be explicitly handled
4489 by this function. Also, the option @option{-fpcc-struct-return}
4490 takes effect regardless of this macro. On most systems, it is
4491 possible to leave the hook undefined; this causes a default
4492 definition to be used, whose value is the constant 1 for @code{BLKmode}
4493 values, and 0 otherwise.
4495 Do not use this hook to indicate that structures and unions should always
4496 be returned in memory. You should instead use @code{DEFAULT_PCC_STRUCT_RETURN}
4500 @defmac DEFAULT_PCC_STRUCT_RETURN
4501 Define this macro to be 1 if all structure and union return values must be
4502 in memory. Since this results in slower code, this should be defined
4503 only if needed for compatibility with other compilers or with an ABI@.
4504 If you define this macro to be 0, then the conventions used for structure
4505 and union return values are decided by the @code{TARGET_RETURN_IN_MEMORY}
4508 If not defined, this defaults to the value 1.
4511 @deftypefn {Target Hook} rtx TARGET_STRUCT_VALUE_RTX (tree @var{fndecl}, int @var{incoming})
4512 This target hook should return the location of the structure value
4513 address (normally a @code{mem} or @code{reg}), or 0 if the address is
4514 passed as an ``invisible'' first argument. Note that @var{fndecl} may
4515 be @code{NULL}, for libcalls. You do not need to define this target
4516 hook if the address is always passed as an ``invisible'' first
4519 On some architectures the place where the structure value address
4520 is found by the called function is not the same place that the
4521 caller put it. This can be due to register windows, or it could
4522 be because the function prologue moves it to a different place.
4523 @var{incoming} is @code{1} or @code{2} when the location is needed in
4524 the context of the called function, and @code{0} in the context of
4527 If @var{incoming} is nonzero and the address is to be found on the
4528 stack, return a @code{mem} which refers to the frame pointer. If
4529 @var{incoming} is @code{2}, the result is being used to fetch the
4530 structure value address at the beginning of a function. If you need
4531 to emit adjusting code, you should do it at this point.
4534 @defmac PCC_STATIC_STRUCT_RETURN
4535 Define this macro if the usual system convention on the target machine
4536 for returning structures and unions is for the called function to return
4537 the address of a static variable containing the value.
4539 Do not define this if the usual system convention is for the caller to
4540 pass an address to the subroutine.
4542 This macro has effect in @option{-fpcc-struct-return} mode, but it does
4543 nothing when you use @option{-freg-struct-return} mode.
4547 @subsection Caller-Saves Register Allocation
4549 If you enable it, GCC can save registers around function calls. This
4550 makes it possible to use call-clobbered registers to hold variables that
4551 must live across calls.
4553 @defmac CALLER_SAVE_PROFITABLE (@var{refs}, @var{calls})
4554 A C expression to determine whether it is worthwhile to consider placing
4555 a pseudo-register in a call-clobbered hard register and saving and
4556 restoring it around each function call. The expression should be 1 when
4557 this is worth doing, and 0 otherwise.
4559 If you don't define this macro, a default is used which is good on most
4560 machines: @code{4 * @var{calls} < @var{refs}}.
4563 @defmac HARD_REGNO_CALLER_SAVE_MODE (@var{regno}, @var{nregs})
4564 A C expression specifying which mode is required for saving @var{nregs}
4565 of a pseudo-register in call-clobbered hard register @var{regno}. If
4566 @var{regno} is unsuitable for caller save, @code{VOIDmode} should be
4567 returned. For most machines this macro need not be defined since GCC
4568 will select the smallest suitable mode.
4571 @node Function Entry
4572 @subsection Function Entry and Exit
4573 @cindex function entry and exit
4577 This section describes the macros that output function entry
4578 (@dfn{prologue}) and exit (@dfn{epilogue}) code.
4580 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_PROLOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4581 If defined, a function that outputs the assembler code for entry to a
4582 function. The prologue is responsible for setting up the stack frame,
4583 initializing the frame pointer register, saving registers that must be
4584 saved, and allocating @var{size} additional bytes of storage for the
4585 local variables. @var{size} is an integer. @var{file} is a stdio
4586 stream to which the assembler code should be output.
4588 The label for the beginning of the function need not be output by this
4589 macro. That has already been done when the macro is run.
4591 @findex regs_ever_live
4592 To determine which registers to save, the macro can refer to the array
4593 @code{regs_ever_live}: element @var{r} is nonzero if hard register
4594 @var{r} is used anywhere within the function. This implies the function
4595 prologue should save register @var{r}, provided it is not one of the
4596 call-used registers. (@code{TARGET_ASM_FUNCTION_EPILOGUE} must likewise use
4597 @code{regs_ever_live}.)
4599 On machines that have ``register windows'', the function entry code does
4600 not save on the stack the registers that are in the windows, even if
4601 they are supposed to be preserved by function calls; instead it takes
4602 appropriate steps to ``push'' the register stack, if any non-call-used
4603 registers are used in the function.
4605 @findex frame_pointer_needed
4606 On machines where functions may or may not have frame-pointers, the
4607 function entry code must vary accordingly; it must set up the frame
4608 pointer if one is wanted, and not otherwise. To determine whether a
4609 frame pointer is in wanted, the macro can refer to the variable
4610 @code{frame_pointer_needed}. The variable's value will be 1 at run
4611 time in a function that needs a frame pointer. @xref{Elimination}.
4613 The function entry code is responsible for allocating any stack space
4614 required for the function. This stack space consists of the regions
4615 listed below. In most cases, these regions are allocated in the
4616 order listed, with the last listed region closest to the top of the
4617 stack (the lowest address if @code{STACK_GROWS_DOWNWARD} is defined, and
4618 the highest address if it is not defined). You can use a different order
4619 for a machine if doing so is more convenient or required for
4620 compatibility reasons. Except in cases where required by standard
4621 or by a debugger, there is no reason why the stack layout used by GCC
4622 need agree with that used by other compilers for a machine.
4625 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_END_PROLOGUE (FILE *@var{file})
4626 If defined, a function that outputs assembler code at the end of a
4627 prologue. This should be used when the function prologue is being
4628 emitted as RTL, and you have some extra assembler that needs to be
4629 emitted. @xref{prologue instruction pattern}.
4632 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_BEGIN_EPILOGUE (FILE *@var{file})
4633 If defined, a function that outputs assembler code at the start of an
4634 epilogue. This should be used when the function epilogue is being
4635 emitted as RTL, and you have some extra assembler that needs to be
4636 emitted. @xref{epilogue instruction pattern}.
4639 @deftypefn {Target Hook} void TARGET_ASM_FUNCTION_EPILOGUE (FILE *@var{file}, HOST_WIDE_INT @var{size})
4640 If defined, a function that outputs the assembler code for exit from a
4641 function. The epilogue is responsible for restoring the saved
4642 registers and stack pointer to their values when the function was
4643 called, and returning control to the caller. This macro takes the
4644 same arguments as the macro @code{TARGET_ASM_FUNCTION_PROLOGUE}, and the
4645 registers to restore are determined from @code{regs_ever_live} and
4646 @code{CALL_USED_REGISTERS} in the same way.
4648 On some machines, there is a single instruction that does all the work
4649 of returning from the function. On these machines, give that
4650 instruction the name @samp{return} and do not define the macro
4651 @code{TARGET_ASM_FUNCTION_EPILOGUE} at all.
4653 Do not define a pattern named @samp{return} if you want the
4654 @code{TARGET_ASM_FUNCTION_EPILOGUE} to be used. If you want the target
4655 switches to control whether return instructions or epilogues are used,
4656 define a @samp{return} pattern with a validity condition that tests the
4657 target switches appropriately. If the @samp{return} pattern's validity
4658 condition is false, epilogues will be used.
4660 On machines where functions may or may not have frame-pointers, the
4661 function exit code must vary accordingly. Sometimes the code for these
4662 two cases is completely different. To determine whether a frame pointer
4663 is wanted, the macro can refer to the variable
4664 @code{frame_pointer_needed}. The variable's value will be 1 when compiling
4665 a function that needs a frame pointer.
4667 Normally, @code{TARGET_ASM_FUNCTION_PROLOGUE} and
4668 @code{TARGET_ASM_FUNCTION_EPILOGUE} must treat leaf functions specially.
4669 The C variable @code{current_function_is_leaf} is nonzero for such a
4670 function. @xref{Leaf Functions}.
4672 On some machines, some functions pop their arguments on exit while
4673 others leave that for the caller to do. For example, the 68020 when
4674 given @option{-mrtd} pops arguments in functions that take a fixed
4675 number of arguments.
4677 @findex current_function_pops_args
4678 Your definition of the macro @code{RETURN_POPS_ARGS} decides which
4679 functions pop their own arguments. @code{TARGET_ASM_FUNCTION_EPILOGUE}
4680 needs to know what was decided. The variable that is called
4681 @code{current_function_pops_args} is the number of bytes of its
4682 arguments that a function should pop. @xref{Scalar Return}.
4683 @c what is the "its arguments" in the above sentence referring to, pray
4684 @c tell? --mew 5feb93
4689 @findex current_function_pretend_args_size
4690 A region of @code{current_function_pretend_args_size} bytes of
4691 uninitialized space just underneath the first argument arriving on the
4692 stack. (This may not be at the very start of the allocated stack region
4693 if the calling sequence has pushed anything else since pushing the stack
4694 arguments. But usually, on such machines, nothing else has been pushed
4695 yet, because the function prologue itself does all the pushing.) This
4696 region is used on machines where an argument may be passed partly in
4697 registers and partly in memory, and, in some cases to support the
4698 features in @code{<stdarg.h>}.
4701 An area of memory used to save certain registers used by the function.
4702 The size of this area, which may also include space for such things as
4703 the return address and pointers to previous stack frames, is
4704 machine-specific and usually depends on which registers have been used
4705 in the function. Machines with register windows often do not require
4709 A region of at least @var{size} bytes, possibly rounded up to an allocation
4710 boundary, to contain the local variables of the function. On some machines,
4711 this region and the save area may occur in the opposite order, with the
4712 save area closer to the top of the stack.
4715 @cindex @code{ACCUMULATE_OUTGOING_ARGS} and stack frames
4716 Optionally, when @code{ACCUMULATE_OUTGOING_ARGS} is defined, a region of
4717 @code{current_function_outgoing_args_size} bytes to be used for outgoing
4718 argument lists of the function. @xref{Stack Arguments}.
4721 @defmac EXIT_IGNORE_STACK
4722 Define this macro as a C expression that is nonzero if the return
4723 instruction or the function epilogue ignores the value of the stack
4724 pointer; in other words, if it is safe to delete an instruction to
4725 adjust the stack pointer before a return from the function. The
4728 Note that this macro's value is relevant only for functions for which
4729 frame pointers are maintained. It is never safe to delete a final
4730 stack adjustment in a function that has no frame pointer, and the
4731 compiler knows this regardless of @code{EXIT_IGNORE_STACK}.
4734 @defmac EPILOGUE_USES (@var{regno})
4735 Define this macro as a C expression that is nonzero for registers that are
4736 used by the epilogue or the @samp{return} pattern. The stack and frame
4737 pointer registers are already assumed to be used as needed.
4740 @defmac EH_USES (@var{regno})
4741 Define this macro as a C expression that is nonzero for registers that are
4742 used by the exception handling mechanism, and so should be considered live
4743 on entry to an exception edge.
4746 @defmac DELAY_SLOTS_FOR_EPILOGUE
4747 Define this macro if the function epilogue contains delay slots to which
4748 instructions from the rest of the function can be ``moved''. The
4749 definition should be a C expression whose value is an integer
4750 representing the number of delay slots there.
4753 @defmac ELIGIBLE_FOR_EPILOGUE_DELAY (@var{insn}, @var{n})
4754 A C expression that returns 1 if @var{insn} can be placed in delay
4755 slot number @var{n} of the epilogue.
4757 The argument @var{n} is an integer which identifies the delay slot now
4758 being considered (since different slots may have different rules of
4759 eligibility). It is never negative and is always less than the number
4760 of epilogue delay slots (what @code{DELAY_SLOTS_FOR_EPILOGUE} returns).
4761 If you reject a particular insn for a given delay slot, in principle, it
4762 may be reconsidered for a subsequent delay slot. Also, other insns may
4763 (at least in principle) be considered for the so far unfilled delay
4766 @findex current_function_epilogue_delay_list
4767 @findex final_scan_insn
4768 The insns accepted to fill the epilogue delay slots are put in an RTL
4769 list made with @code{insn_list} objects, stored in the variable
4770 @code{current_function_epilogue_delay_list}. The insn for the first
4771 delay slot comes first in the list. Your definition of the macro
4772 @code{TARGET_ASM_FUNCTION_EPILOGUE} should fill the delay slots by
4773 outputting the insns in this list, usually by calling
4774 @code{final_scan_insn}.
4776 You need not define this macro if you did not define
4777 @code{DELAY_SLOTS_FOR_EPILOGUE}.
4780 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_MI_THUNK (FILE *@var{file}, tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4781 A function that outputs the assembler code for a thunk
4782 function, used to implement C++ virtual function calls with multiple
4783 inheritance. The thunk acts as a wrapper around a virtual function,
4784 adjusting the implicit object parameter before handing control off to
4787 First, emit code to add the integer @var{delta} to the location that
4788 contains the incoming first argument. Assume that this argument
4789 contains a pointer, and is the one used to pass the @code{this} pointer
4790 in C++. This is the incoming argument @emph{before} the function prologue,
4791 e.g.@: @samp{%o0} on a sparc. The addition must preserve the values of
4792 all other incoming arguments.
4794 Then, if @var{vcall_offset} is nonzero, an additional adjustment should be
4795 made after adding @code{delta}. In particular, if @var{p} is the
4796 adjusted pointer, the following adjustment should be made:
4799 p += (*((ptrdiff_t **)p))[vcall_offset/sizeof(ptrdiff_t)]
4802 After the additions, emit code to jump to @var{function}, which is a
4803 @code{FUNCTION_DECL}. This is a direct pure jump, not a call, and does
4804 not touch the return address. Hence returning from @var{FUNCTION} will
4805 return to whoever called the current @samp{thunk}.
4807 The effect must be as if @var{function} had been called directly with
4808 the adjusted first argument. This macro is responsible for emitting all
4809 of the code for a thunk function; @code{TARGET_ASM_FUNCTION_PROLOGUE}
4810 and @code{TARGET_ASM_FUNCTION_EPILOGUE} are not invoked.
4812 The @var{thunk_fndecl} is redundant. (@var{delta} and @var{function}
4813 have already been extracted from it.) It might possibly be useful on
4814 some targets, but probably not.
4816 If you do not define this macro, the target-independent code in the C++
4817 front end will generate a less efficient heavyweight thunk that calls
4818 @var{function} instead of jumping to it. The generic approach does
4819 not support varargs.
4822 @deftypefn {Target Hook} bool TARGET_ASM_CAN_OUTPUT_MI_THUNK (tree @var{thunk_fndecl}, HOST_WIDE_INT @var{delta}, HOST_WIDE_INT @var{vcall_offset}, tree @var{function})
4823 A function that returns true if TARGET_ASM_OUTPUT_MI_THUNK would be able
4824 to output the assembler code for the thunk function specified by the
4825 arguments it is passed, and false otherwise. In the latter case, the
4826 generic approach will be used by the C++ front end, with the limitations
4831 @subsection Generating Code for Profiling
4832 @cindex profiling, code generation
4834 These macros will help you generate code for profiling.
4836 @defmac FUNCTION_PROFILER (@var{file}, @var{labelno})
4837 A C statement or compound statement to output to @var{file} some
4838 assembler code to call the profiling subroutine @code{mcount}.
4841 The details of how @code{mcount} expects to be called are determined by
4842 your operating system environment, not by GCC@. To figure them out,
4843 compile a small program for profiling using the system's installed C
4844 compiler and look at the assembler code that results.
4846 Older implementations of @code{mcount} expect the address of a counter
4847 variable to be loaded into some register. The name of this variable is
4848 @samp{LP} followed by the number @var{labelno}, so you would generate
4849 the name using @samp{LP%d} in a @code{fprintf}.
4852 @defmac PROFILE_HOOK
4853 A C statement or compound statement to output to @var{file} some assembly
4854 code to call the profiling subroutine @code{mcount} even the target does
4855 not support profiling.
4858 @defmac NO_PROFILE_COUNTERS
4859 Define this macro to be an expression with a nonzero value if the
4860 @code{mcount} subroutine on your system does not need a counter variable
4861 allocated for each function. This is true for almost all modern
4862 implementations. If you define this macro, you must not use the
4863 @var{labelno} argument to @code{FUNCTION_PROFILER}.
4866 @defmac PROFILE_BEFORE_PROLOGUE
4867 Define this macro if the code for function profiling should come before
4868 the function prologue. Normally, the profiling code comes after.
4872 @subsection Permitting tail calls
4875 @deftypefn {Target Hook} bool TARGET_FUNCTION_OK_FOR_SIBCALL (tree @var{decl}, tree @var{exp})
4876 True if it is ok to do sibling call optimization for the specified
4877 call expression @var{exp}. @var{decl} will be the called function,
4878 or @code{NULL} if this is an indirect call.
4880 It is not uncommon for limitations of calling conventions to prevent
4881 tail calls to functions outside the current unit of translation, or
4882 during PIC compilation. The hook is used to enforce these restrictions,
4883 as the @code{sibcall} md pattern can not fail, or fall over to a
4884 ``normal'' call. The criteria for successful sibling call optimization
4885 may vary greatly between different architectures.
4888 @deftypefn {Target Hook} void TARGET_EXTRA_LIVE_ON_ENTRY (bitmap *@var{regs})
4889 Add any hard registers to @var{regs} that are live on entry to the
4890 function. This hook only needs to be defined to provide registers that
4891 cannot be found by examination of FUNCTION_ARG_REGNO_P, the callee saved
4892 registers, STATIC_CHAIN_INCOMING_REGNUM, STATIC_CHAIN_REGNUM,
4893 TARGET_STRUCT_VALUE_RTX, FRAME_POINTER_REGNUM, EH_USES,
4894 FRAME_POINTER_REGNUM, ARG_POINTER_REGNUM, and the PIC_OFFSET_TABLE_REGNUM.
4897 @node Stack Smashing Protection
4898 @subsection Stack smashing protection
4899 @cindex stack smashing protection
4901 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_GUARD (void)
4902 This hook returns a @code{DECL} node for the external variable to use
4903 for the stack protection guard. This variable is initialized by the
4904 runtime to some random value and is used to initialize the guard value
4905 that is placed at the top of the local stack frame. The type of this
4906 variable must be @code{ptr_type_node}.
4908 The default version of this hook creates a variable called
4909 @samp{__stack_chk_guard}, which is normally defined in @file{libgcc2.c}.
4912 @deftypefn {Target Hook} tree TARGET_STACK_PROTECT_FAIL (void)
4913 This hook returns a tree expression that alerts the runtime that the
4914 stack protect guard variable has been modified. This expression should
4915 involve a call to a @code{noreturn} function.
4917 The default version of this hook invokes a function called
4918 @samp{__stack_chk_fail}, taking no arguments. This function is
4919 normally defined in @file{libgcc2.c}.
4923 @section Implementing the Varargs Macros
4924 @cindex varargs implementation
4926 GCC comes with an implementation of @code{<varargs.h>} and
4927 @code{<stdarg.h>} that work without change on machines that pass arguments
4928 on the stack. Other machines require their own implementations of
4929 varargs, and the two machine independent header files must have
4930 conditionals to include it.
4932 ISO @code{<stdarg.h>} differs from traditional @code{<varargs.h>} mainly in
4933 the calling convention for @code{va_start}. The traditional
4934 implementation takes just one argument, which is the variable in which
4935 to store the argument pointer. The ISO implementation of
4936 @code{va_start} takes an additional second argument. The user is
4937 supposed to write the last named argument of the function here.
4939 However, @code{va_start} should not use this argument. The way to find
4940 the end of the named arguments is with the built-in functions described
4943 @defmac __builtin_saveregs ()
4944 Use this built-in function to save the argument registers in memory so
4945 that the varargs mechanism can access them. Both ISO and traditional
4946 versions of @code{va_start} must use @code{__builtin_saveregs}, unless
4947 you use @code{TARGET_SETUP_INCOMING_VARARGS} (see below) instead.
4949 On some machines, @code{__builtin_saveregs} is open-coded under the
4950 control of the target hook @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. On
4951 other machines, it calls a routine written in assembler language,
4952 found in @file{libgcc2.c}.
4954 Code generated for the call to @code{__builtin_saveregs} appears at the
4955 beginning of the function, as opposed to where the call to
4956 @code{__builtin_saveregs} is written, regardless of what the code is.
4957 This is because the registers must be saved before the function starts
4958 to use them for its own purposes.
4959 @c i rewrote the first sentence above to fix an overfull hbox. --mew
4963 @defmac __builtin_args_info (@var{category})
4964 Use this built-in function to find the first anonymous arguments in
4967 In general, a machine may have several categories of registers used for
4968 arguments, each for a particular category of data types. (For example,
4969 on some machines, floating-point registers are used for floating-point
4970 arguments while other arguments are passed in the general registers.)
4971 To make non-varargs functions use the proper calling convention, you
4972 have defined the @code{CUMULATIVE_ARGS} data type to record how many
4973 registers in each category have been used so far
4975 @code{__builtin_args_info} accesses the same data structure of type
4976 @code{CUMULATIVE_ARGS} after the ordinary argument layout is finished
4977 with it, with @var{category} specifying which word to access. Thus, the
4978 value indicates the first unused register in a given category.
4980 Normally, you would use @code{__builtin_args_info} in the implementation
4981 of @code{va_start}, accessing each category just once and storing the
4982 value in the @code{va_list} object. This is because @code{va_list} will
4983 have to update the values, and there is no way to alter the
4984 values accessed by @code{__builtin_args_info}.
4987 @defmac __builtin_next_arg (@var{lastarg})
4988 This is the equivalent of @code{__builtin_args_info}, for stack
4989 arguments. It returns the address of the first anonymous stack
4990 argument, as type @code{void *}. If @code{ARGS_GROW_DOWNWARD}, it
4991 returns the address of the location above the first anonymous stack
4992 argument. Use it in @code{va_start} to initialize the pointer for
4993 fetching arguments from the stack. Also use it in @code{va_start} to
4994 verify that the second parameter @var{lastarg} is the last named argument
4995 of the current function.
4998 @defmac __builtin_classify_type (@var{object})
4999 Since each machine has its own conventions for which data types are
5000 passed in which kind of register, your implementation of @code{va_arg}
5001 has to embody these conventions. The easiest way to categorize the
5002 specified data type is to use @code{__builtin_classify_type} together
5003 with @code{sizeof} and @code{__alignof__}.
5005 @code{__builtin_classify_type} ignores the value of @var{object},
5006 considering only its data type. It returns an integer describing what
5007 kind of type that is---integer, floating, pointer, structure, and so on.
5009 The file @file{typeclass.h} defines an enumeration that you can use to
5010 interpret the values of @code{__builtin_classify_type}.
5013 These machine description macros help implement varargs:
5015 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN_SAVEREGS (void)
5016 If defined, this hook produces the machine-specific code for a call to
5017 @code{__builtin_saveregs}. This code will be moved to the very
5018 beginning of the function, before any parameter access are made. The
5019 return value of this function should be an RTX that contains the value
5020 to use as the return of @code{__builtin_saveregs}.
5023 @deftypefn {Target Hook} void TARGET_SETUP_INCOMING_VARARGS (CUMULATIVE_ARGS *@var{args_so_far}, enum machine_mode @var{mode}, tree @var{type}, int *@var{pretend_args_size}, int @var{second_time})
5024 This target hook offers an alternative to using
5025 @code{__builtin_saveregs} and defining the hook
5026 @code{TARGET_EXPAND_BUILTIN_SAVEREGS}. Use it to store the anonymous
5027 register arguments into the stack so that all the arguments appear to
5028 have been passed consecutively on the stack. Once this is done, you can
5029 use the standard implementation of varargs that works for machines that
5030 pass all their arguments on the stack.
5032 The argument @var{args_so_far} points to the @code{CUMULATIVE_ARGS} data
5033 structure, containing the values that are obtained after processing the
5034 named arguments. The arguments @var{mode} and @var{type} describe the
5035 last named argument---its machine mode and its data type as a tree node.
5037 The target hook should do two things: first, push onto the stack all the
5038 argument registers @emph{not} used for the named arguments, and second,
5039 store the size of the data thus pushed into the @code{int}-valued
5040 variable pointed to by @var{pretend_args_size}. The value that you
5041 store here will serve as additional offset for setting up the stack
5044 Because you must generate code to push the anonymous arguments at
5045 compile time without knowing their data types,
5046 @code{TARGET_SETUP_INCOMING_VARARGS} is only useful on machines that
5047 have just a single category of argument register and use it uniformly
5050 If the argument @var{second_time} is nonzero, it means that the
5051 arguments of the function are being analyzed for the second time. This
5052 happens for an inline function, which is not actually compiled until the
5053 end of the source file. The hook @code{TARGET_SETUP_INCOMING_VARARGS} should
5054 not generate any instructions in this case.
5057 @deftypefn {Target Hook} bool TARGET_STRICT_ARGUMENT_NAMING (CUMULATIVE_ARGS *@var{ca})
5058 Define this hook to return @code{true} if the location where a function
5059 argument is passed depends on whether or not it is a named argument.
5061 This hook controls how the @var{named} argument to @code{FUNCTION_ARG}
5062 is set for varargs and stdarg functions. If this hook returns
5063 @code{true}, the @var{named} argument is always true for named
5064 arguments, and false for unnamed arguments. If it returns @code{false},
5065 but @code{TARGET_PRETEND_OUTGOING_VARARGS_NAMED} returns @code{true},
5066 then all arguments are treated as named. Otherwise, all named arguments
5067 except the last are treated as named.
5069 You need not define this hook if it always returns zero.
5072 @deftypefn {Target Hook} bool TARGET_PRETEND_OUTGOING_VARARGS_NAMED
5073 If you need to conditionally change ABIs so that one works with
5074 @code{TARGET_SETUP_INCOMING_VARARGS}, but the other works like neither
5075 @code{TARGET_SETUP_INCOMING_VARARGS} nor @code{TARGET_STRICT_ARGUMENT_NAMING} was
5076 defined, then define this hook to return @code{true} if
5077 @code{TARGET_SETUP_INCOMING_VARARGS} is used, @code{false} otherwise.
5078 Otherwise, you should not define this hook.
5082 @section Trampolines for Nested Functions
5083 @cindex trampolines for nested functions
5084 @cindex nested functions, trampolines for
5086 A @dfn{trampoline} is a small piece of code that is created at run time
5087 when the address of a nested function is taken. It normally resides on
5088 the stack, in the stack frame of the containing function. These macros
5089 tell GCC how to generate code to allocate and initialize a
5092 The instructions in the trampoline must do two things: load a constant
5093 address into the static chain register, and jump to the real address of
5094 the nested function. On CISC machines such as the m68k, this requires
5095 two instructions, a move immediate and a jump. Then the two addresses
5096 exist in the trampoline as word-long immediate operands. On RISC
5097 machines, it is often necessary to load each address into a register in
5098 two parts. Then pieces of each address form separate immediate
5101 The code generated to initialize the trampoline must store the variable
5102 parts---the static chain value and the function address---into the
5103 immediate operands of the instructions. On a CISC machine, this is
5104 simply a matter of copying each address to a memory reference at the
5105 proper offset from the start of the trampoline. On a RISC machine, it
5106 may be necessary to take out pieces of the address and store them
5109 @deftypefn {Target Hook} void TARGET_ASM_TRAMPOLINE_TEMPLATE (FILE *@var{f})
5110 This hook is called by @code{assemble_trampoline_template} to output,
5111 on the stream @var{f}, assembler code for a block of data that contains
5112 the constant parts of a trampoline. This code should not include a
5113 label---the label is taken care of automatically.
5115 If you do not define this hook, it means no template is needed
5116 for the target. Do not define this hook on systems where the block move
5117 code to copy the trampoline into place would be larger than the code
5118 to generate it on the spot.
5121 @defmac TRAMPOLINE_SECTION
5122 Return the section into which the trampoline template is to be placed
5123 (@pxref{Sections}). The default value is @code{readonly_data_section}.
5126 @defmac TRAMPOLINE_SIZE
5127 A C expression for the size in bytes of the trampoline, as an integer.
5130 @defmac TRAMPOLINE_ALIGNMENT
5131 Alignment required for trampolines, in bits.
5133 If you don't define this macro, the value of @code{FUNCTION_ALIGNMENT}
5134 is used for aligning trampolines.
5137 @deftypefn {Target Hook} void TARGET_TRAMPOLINE_INIT (rtx @var{m_tramp}, tree @var{fndecl}, rtx @var{static_chain})
5138 This hook is called to initialize a trampoline.
5139 @var{m_tramp} is an RTX for the memory block for the trampoline; @var{fndecl}
5140 is the @code{FUNCTION_DECL} for the nested function; @var{static_chain} is an
5141 RTX for the static chain value that should be passed to the function
5144 If the target defines @code{TARGET_ASM_TRAMPOLINE_TEMPLATE}, then the
5145 first thing this hook should do is emit a block move into @var{m_tramp}
5146 from the memory block returned by @code{assemble_trampoline_template}.
5147 Note that the block move need only cover the constant parts of the
5148 trampoline. If the target isolates the variable parts of the trampoline
5149 to the end, not all @code{TRAMPOLINE_SIZE} bytes need be copied.
5151 If the target requires any other actions, such as flushing caches or
5152 enabling stack execution, these actions should be performed after
5153 initializing the trampoline proper.
5156 @deftypefn {Target Hook} rtx TARGET_TRAMPOLINE_ADJUST_ADDRESS (rtx @var{addr})
5157 This hook should perform any machine-specific adjustment in
5158 the address of the trampoline. Its argument contains the address of the
5159 memory block that was passed to @code{TARGET_TRAMPOLINE_INIT}. In case
5160 the address to be used for a function call should be different from the
5161 address at which the template was stored, the different address should
5162 be returned; otherwise @var{addr} should be returned unchanged.
5163 If this hook is not defined, @var{addr} will be used for function calls.
5166 Implementing trampolines is difficult on many machines because they have
5167 separate instruction and data caches. Writing into a stack location
5168 fails to clear the memory in the instruction cache, so when the program
5169 jumps to that location, it executes the old contents.
5171 Here are two possible solutions. One is to clear the relevant parts of
5172 the instruction cache whenever a trampoline is set up. The other is to
5173 make all trampolines identical, by having them jump to a standard
5174 subroutine. The former technique makes trampoline execution faster; the
5175 latter makes initialization faster.
5177 To clear the instruction cache when a trampoline is initialized, define
5178 the following macro.
5180 @defmac CLEAR_INSN_CACHE (@var{beg}, @var{end})
5181 If defined, expands to a C expression clearing the @emph{instruction
5182 cache} in the specified interval. The definition of this macro would
5183 typically be a series of @code{asm} statements. Both @var{beg} and
5184 @var{end} are both pointer expressions.
5187 The operating system may also require the stack to be made executable
5188 before calling the trampoline. To implement this requirement, define
5189 the following macro.
5191 @defmac ENABLE_EXECUTE_STACK
5192 Define this macro if certain operations must be performed before executing
5193 code located on the stack. The macro should expand to a series of C
5194 file-scope constructs (e.g.@: functions) and provide a unique entry point
5195 named @code{__enable_execute_stack}. The target is responsible for
5196 emitting calls to the entry point in the code, for example from the
5197 @code{TARGET_TRAMPOLINE_INIT} hook.
5200 To use a standard subroutine, define the following macro. In addition,
5201 you must make sure that the instructions in a trampoline fill an entire
5202 cache line with identical instructions, or else ensure that the
5203 beginning of the trampoline code is always aligned at the same point in
5204 its cache line. Look in @file{m68k.h} as a guide.
5206 @defmac TRANSFER_FROM_TRAMPOLINE
5207 Define this macro if trampolines need a special subroutine to do their
5208 work. The macro should expand to a series of @code{asm} statements
5209 which will be compiled with GCC@. They go in a library function named
5210 @code{__transfer_from_trampoline}.
5212 If you need to avoid executing the ordinary prologue code of a compiled
5213 C function when you jump to the subroutine, you can do so by placing a
5214 special label of your own in the assembler code. Use one @code{asm}
5215 statement to generate an assembler label, and another to make the label
5216 global. Then trampolines can use that label to jump directly to your
5217 special assembler code.
5221 @section Implicit Calls to Library Routines
5222 @cindex library subroutine names
5223 @cindex @file{libgcc.a}
5225 @c prevent bad page break with this line
5226 Here is an explanation of implicit calls to library routines.
5228 @defmac DECLARE_LIBRARY_RENAMES
5229 This macro, if defined, should expand to a piece of C code that will get
5230 expanded when compiling functions for libgcc.a. It can be used to
5231 provide alternate names for GCC's internal library functions if there
5232 are ABI-mandated names that the compiler should provide.
5235 @findex init_one_libfunc
5236 @findex set_optab_libfunc
5237 @deftypefn {Target Hook} void TARGET_INIT_LIBFUNCS (void)
5238 This hook should declare additional library routines or rename
5239 existing ones, using the functions @code{set_optab_libfunc} and
5240 @code{init_one_libfunc} defined in @file{optabs.c}.
5241 @code{init_optabs} calls this macro after initializing all the normal
5244 The default is to do nothing. Most ports don't need to define this hook.
5247 @defmac FLOAT_LIB_COMPARE_RETURNS_BOOL (@var{mode}, @var{comparison})
5248 This macro should return @code{true} if the library routine that
5249 implements the floating point comparison operator @var{comparison} in
5250 mode @var{mode} will return a boolean, and @var{false} if it will
5253 GCC's own floating point libraries return tristates from the
5254 comparison operators, so the default returns false always. Most ports
5255 don't need to define this macro.
5258 @defmac TARGET_LIB_INT_CMP_BIASED
5259 This macro should evaluate to @code{true} if the integer comparison
5260 functions (like @code{__cmpdi2}) return 0 to indicate that the first
5261 operand is smaller than the second, 1 to indicate that they are equal,
5262 and 2 to indicate that the first operand is greater than the second.
5263 If this macro evaluates to @code{false} the comparison functions return
5264 @minus{}1, 0, and 1 instead of 0, 1, and 2. If the target uses the routines
5265 in @file{libgcc.a}, you do not need to define this macro.
5268 @cindex US Software GOFAST, floating point emulation library
5269 @cindex floating point emulation library, US Software GOFAST
5270 @cindex GOFAST, floating point emulation library
5271 @findex gofast_maybe_init_libfuncs
5272 @defmac US_SOFTWARE_GOFAST
5273 Define this macro if your system C library uses the US Software GOFAST
5274 library to provide floating point emulation.
5276 In addition to defining this macro, your architecture must set
5277 @code{TARGET_INIT_LIBFUNCS} to @code{gofast_maybe_init_libfuncs}, or
5278 else call that function from its version of that hook. It is defined
5279 in @file{config/gofast.h}, which must be included by your
5280 architecture's @file{@var{cpu}.c} file. See @file{sparc/sparc.c} for
5283 If this macro is defined, the
5284 @code{TARGET_FLOAT_LIB_COMPARE_RETURNS_BOOL} target hook must return
5285 false for @code{SFmode} and @code{DFmode} comparisons.
5288 @cindex @code{EDOM}, implicit usage
5291 The value of @code{EDOM} on the target machine, as a C integer constant
5292 expression. If you don't define this macro, GCC does not attempt to
5293 deposit the value of @code{EDOM} into @code{errno} directly. Look in
5294 @file{/usr/include/errno.h} to find the value of @code{EDOM} on your
5297 If you do not define @code{TARGET_EDOM}, then compiled code reports
5298 domain errors by calling the library function and letting it report the
5299 error. If mathematical functions on your system use @code{matherr} when
5300 there is an error, then you should leave @code{TARGET_EDOM} undefined so
5301 that @code{matherr} is used normally.
5304 @cindex @code{errno}, implicit usage
5305 @defmac GEN_ERRNO_RTX
5306 Define this macro as a C expression to create an rtl expression that
5307 refers to the global ``variable'' @code{errno}. (On certain systems,
5308 @code{errno} may not actually be a variable.) If you don't define this
5309 macro, a reasonable default is used.
5312 @cindex C99 math functions, implicit usage
5313 @defmac TARGET_C99_FUNCTIONS
5314 When this macro is nonzero, GCC will implicitly optimize @code{sin} calls into
5315 @code{sinf} and similarly for other functions defined by C99 standard. The
5316 default is zero because a number of existing systems lack support for these
5317 functions in their runtime so this macro needs to be redefined to one on
5318 systems that do support the C99 runtime.
5321 @cindex sincos math function, implicit usage
5322 @defmac TARGET_HAS_SINCOS
5323 When this macro is nonzero, GCC will implicitly optimize calls to @code{sin}
5324 and @code{cos} with the same argument to a call to @code{sincos}. The
5325 default is zero. The target has to provide the following functions:
5327 void sincos(double x, double *sin, double *cos);
5328 void sincosf(float x, float *sin, float *cos);
5329 void sincosl(long double x, long double *sin, long double *cos);
5333 @defmac NEXT_OBJC_RUNTIME
5334 Define this macro to generate code for Objective-C message sending using
5335 the calling convention of the NeXT system. This calling convention
5336 involves passing the object, the selector and the method arguments all
5337 at once to the method-lookup library function.
5339 The default calling convention passes just the object and the selector
5340 to the lookup function, which returns a pointer to the method.
5343 @node Addressing Modes
5344 @section Addressing Modes
5345 @cindex addressing modes
5347 @c prevent bad page break with this line
5348 This is about addressing modes.
5350 @defmac HAVE_PRE_INCREMENT
5351 @defmacx HAVE_PRE_DECREMENT
5352 @defmacx HAVE_POST_INCREMENT
5353 @defmacx HAVE_POST_DECREMENT
5354 A C expression that is nonzero if the machine supports pre-increment,
5355 pre-decrement, post-increment, or post-decrement addressing respectively.
5358 @defmac HAVE_PRE_MODIFY_DISP
5359 @defmacx HAVE_POST_MODIFY_DISP
5360 A C expression that is nonzero if the machine supports pre- or
5361 post-address side-effect generation involving constants other than
5362 the size of the memory operand.
5365 @defmac HAVE_PRE_MODIFY_REG
5366 @defmacx HAVE_POST_MODIFY_REG
5367 A C expression that is nonzero if the machine supports pre- or
5368 post-address side-effect generation involving a register displacement.
5371 @defmac CONSTANT_ADDRESS_P (@var{x})
5372 A C expression that is 1 if the RTX @var{x} is a constant which
5373 is a valid address. On most machines, this can be defined as
5374 @code{CONSTANT_P (@var{x})}, but a few machines are more restrictive
5375 in which constant addresses are supported.
5378 @defmac CONSTANT_P (@var{x})
5379 @code{CONSTANT_P}, which is defined by target-independent code,
5380 accepts integer-values expressions whose values are not explicitly
5381 known, such as @code{symbol_ref}, @code{label_ref}, and @code{high}
5382 expressions and @code{const} arithmetic expressions, in addition to
5383 @code{const_int} and @code{const_double} expressions.
5386 @defmac MAX_REGS_PER_ADDRESS
5387 A number, the maximum number of registers that can appear in a valid
5388 memory address. Note that it is up to you to specify a value equal to
5389 the maximum number that @code{TARGET_LEGITIMATE_ADDRESS_P} would ever
5393 @deftypefn {Target Hook} TARGET_LEGITIMATE_ADDRESS_P (enum machine_mode @var{mode}, rtx @var{x}, bool @var{strict})
5394 A function that returns whether @var{x} (an RTX) is a legitimate memory
5395 address on the target machine for a memory operand of mode @var{mode}.
5397 Legitimate addresses are defined in two variants: a strict variant and a
5398 non-strict one. The @code{strict} parameter chooses which variant is
5399 desired by the caller.
5401 The strict variant is used in the reload pass. It must be defined so
5402 that any pseudo-register that has not been allocated a hard register is
5403 considered a memory reference. This is because in contexts where some
5404 kind of register is required, a pseudo-register with no hard register
5405 must be rejected. For non-hard registers, the strict variant should look
5406 up the @code{reg_renumber} array; it should then proceed using the hard
5407 register number in the array, or treat the pseudo as a memory reference
5408 if the array holds @code{-1}.
5410 The non-strict variant is used in other passes. It must be defined to
5411 accept all pseudo-registers in every context where some kind of
5412 register is required.
5414 Normally, constant addresses which are the sum of a @code{symbol_ref}
5415 and an integer are stored inside a @code{const} RTX to mark them as
5416 constant. Therefore, there is no need to recognize such sums
5417 specifically as legitimate addresses. Normally you would simply
5418 recognize any @code{const} as legitimate.
5420 Usually @code{PRINT_OPERAND_ADDRESS} is not prepared to handle constant
5421 sums that are not marked with @code{const}. It assumes that a naked
5422 @code{plus} indicates indexing. If so, then you @emph{must} reject such
5423 naked constant sums as illegitimate addresses, so that none of them will
5424 be given to @code{PRINT_OPERAND_ADDRESS}.
5426 @cindex @code{TARGET_ENCODE_SECTION_INFO} and address validation
5427 On some machines, whether a symbolic address is legitimate depends on
5428 the section that the address refers to. On these machines, define the
5429 target hook @code{TARGET_ENCODE_SECTION_INFO} to store the information
5430 into the @code{symbol_ref}, and then check for it here. When you see a
5431 @code{const}, you will have to look inside it to find the
5432 @code{symbol_ref} in order to determine the section. @xref{Assembler
5435 @cindex @code{GO_IF_LEGITIMATE_ADDRESS}
5436 Some ports are still using a deprecated legacy substitute for
5437 this hook, the @code{GO_IF_LEGITIMATE_ADDRESS} macro. This macro
5441 #define GO_IF_LEGITIMATE_ADDRESS (@var{mode}, @var{x}, @var{label})
5445 and should @code{goto @var{label}} if the address @var{x} is a valid
5446 address on the target machine for a memory operand of mode @var{mode}.
5447 Whether the strict or non-strict variants are desired is defined by
5448 the @code{REG_OK_STRICT} macro introduced earlier in this section.
5449 Using the hook is usually simpler because it limits the number of
5450 files that are recompiled when changes are made.
5453 @defmac TARGET_MEM_CONSTRAINT
5454 A single character to be used instead of the default @code{'m'}
5455 character for general memory addresses. This defines the constraint
5456 letter which matches the memory addresses accepted by
5457 @code{TARGET_LEGITIMATE_ADDRESS_P}. Define this macro if you want to
5458 support new address formats in your back end without changing the
5459 semantics of the @code{'m'} constraint. This is necessary in order to
5460 preserve functionality of inline assembly constructs using the
5461 @code{'m'} constraint.
5464 @defmac FIND_BASE_TERM (@var{x})
5465 A C expression to determine the base term of address @var{x},
5466 or to provide a simplified version of @var{x} from which @file{alias.c}
5467 can easily find the base term. This macro is used in only two places:
5468 @code{find_base_value} and @code{find_base_term} in @file{alias.c}.
5470 It is always safe for this macro to not be defined. It exists so
5471 that alias analysis can understand machine-dependent addresses.
5473 The typical use of this macro is to handle addresses containing
5474 a label_ref or symbol_ref within an UNSPEC@.
5477 @deftypefn {Target Hook} rtx TARGET_LEGITIMIZE_ADDRESS (rtx @var{x}, rtx @var{oldx}, enum machine_mode @var{mode})
5478 This hook is given an invalid memory address @var{x} for an
5479 operand of mode @var{mode} and should try to return a valid memory
5482 @findex break_out_memory_refs
5483 @var{x} will always be the result of a call to @code{break_out_memory_refs},
5484 and @var{oldx} will be the operand that was given to that function to produce
5487 The code of the hook should not alter the substructure of
5488 @var{x}. If it transforms @var{x} into a more legitimate form, it
5489 should return the new @var{x}.
5491 It is not necessary for this hook to come up with a legitimate address.
5492 The compiler has standard ways of doing so in all cases. In fact, it
5493 is safe to omit this hook or make it return @var{x} if it cannot find
5494 a valid way to legitimize the address. But often a machine-dependent
5495 strategy can generate better code.
5498 @defmac LEGITIMIZE_RELOAD_ADDRESS (@var{x}, @var{mode}, @var{opnum}, @var{type}, @var{ind_levels}, @var{win})
5499 A C compound statement that attempts to replace @var{x}, which is an address
5500 that needs reloading, with a valid memory address for an operand of mode
5501 @var{mode}. @var{win} will be a C statement label elsewhere in the code.
5502 It is not necessary to define this macro, but it might be useful for
5503 performance reasons.
5505 For example, on the i386, it is sometimes possible to use a single
5506 reload register instead of two by reloading a sum of two pseudo
5507 registers into a register. On the other hand, for number of RISC
5508 processors offsets are limited so that often an intermediate address
5509 needs to be generated in order to address a stack slot. By defining
5510 @code{LEGITIMIZE_RELOAD_ADDRESS} appropriately, the intermediate addresses
5511 generated for adjacent some stack slots can be made identical, and thus
5514 @emph{Note}: This macro should be used with caution. It is necessary
5515 to know something of how reload works in order to effectively use this,
5516 and it is quite easy to produce macros that build in too much knowledge
5517 of reload internals.
5519 @emph{Note}: This macro must be able to reload an address created by a
5520 previous invocation of this macro. If it fails to handle such addresses
5521 then the compiler may generate incorrect code or abort.
5524 The macro definition should use @code{push_reload} to indicate parts that
5525 need reloading; @var{opnum}, @var{type} and @var{ind_levels} are usually
5526 suitable to be passed unaltered to @code{push_reload}.
5528 The code generated by this macro must not alter the substructure of
5529 @var{x}. If it transforms @var{x} into a more legitimate form, it
5530 should assign @var{x} (which will always be a C variable) a new value.
5531 This also applies to parts that you change indirectly by calling
5534 @findex strict_memory_address_p
5535 The macro definition may use @code{strict_memory_address_p} to test if
5536 the address has become legitimate.
5539 If you want to change only a part of @var{x}, one standard way of doing
5540 this is to use @code{copy_rtx}. Note, however, that it unshares only a
5541 single level of rtl. Thus, if the part to be changed is not at the
5542 top level, you'll need to replace first the top level.
5543 It is not necessary for this macro to come up with a legitimate
5544 address; but often a machine-dependent strategy can generate better code.
5547 @defmac GO_IF_MODE_DEPENDENT_ADDRESS (@var{addr}, @var{label})
5548 A C statement or compound statement with a conditional @code{goto
5549 @var{label};} executed if memory address @var{x} (an RTX) can have
5550 different meanings depending on the machine mode of the memory
5551 reference it is used for or if the address is valid for some modes
5554 Autoincrement and autodecrement addresses typically have mode-dependent
5555 effects because the amount of the increment or decrement is the size
5556 of the operand being addressed. Some machines have other mode-dependent
5557 addresses. Many RISC machines have no mode-dependent addresses.
5559 You may assume that @var{addr} is a valid address for the machine.
5562 @defmac LEGITIMATE_CONSTANT_P (@var{x})
5563 A C expression that is nonzero if @var{x} is a legitimate constant for
5564 an immediate operand on the target machine. You can assume that
5565 @var{x} satisfies @code{CONSTANT_P}, so you need not check this. In fact,
5566 @samp{1} is a suitable definition for this macro on machines where
5567 anything @code{CONSTANT_P} is valid.
5570 @deftypefn {Target Hook} rtx TARGET_DELEGITIMIZE_ADDRESS (rtx @var{x})
5571 This hook is used to undo the possibly obfuscating effects of the
5572 @code{LEGITIMIZE_ADDRESS} and @code{LEGITIMIZE_RELOAD_ADDRESS} target
5573 macros. Some backend implementations of these macros wrap symbol
5574 references inside an @code{UNSPEC} rtx to represent PIC or similar
5575 addressing modes. This target hook allows GCC's optimizers to understand
5576 the semantics of these opaque @code{UNSPEC}s by converting them back
5577 into their original form.
5580 @deftypefn {Target Hook} bool TARGET_CANNOT_FORCE_CONST_MEM (rtx @var{x})
5581 This hook should return true if @var{x} is of a form that cannot (or
5582 should not) be spilled to the constant pool. The default version of
5583 this hook returns false.
5585 The primary reason to define this hook is to prevent reload from
5586 deciding that a non-legitimate constant would be better reloaded
5587 from the constant pool instead of spilling and reloading a register
5588 holding the constant. This restriction is often true of addresses
5589 of TLS symbols for various targets.
5592 @deftypefn {Target Hook} bool TARGET_USE_BLOCKS_FOR_CONSTANT_P (enum machine_mode @var{mode}, rtx @var{x})
5593 This hook should return true if pool entries for constant @var{x} can
5594 be placed in an @code{object_block} structure. @var{mode} is the mode
5597 The default version returns false for all constants.
5600 @deftypefn {Target Hook} tree TARGET_BUILTIN_RECIPROCAL (enum tree_code @var{fn}, bool @var{tm_fn}, bool @var{sqrt})
5601 This hook should return the DECL of a function that implements reciprocal of
5602 the builtin function with builtin function code @var{fn}, or
5603 @code{NULL_TREE} if such a function is not available. @var{tm_fn} is true
5604 when @var{fn} is a code of a machine-dependent builtin function. When
5605 @var{sqrt} is true, additional optimizations that apply only to the reciprocal
5606 of a square root function are performed, and only reciprocals of @code{sqrt}
5610 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MASK_FOR_LOAD (void)
5611 This hook should return the DECL of a function @var{f} that given an
5612 address @var{addr} as an argument returns a mask @var{m} that can be
5613 used to extract from two vectors the relevant data that resides in
5614 @var{addr} in case @var{addr} is not properly aligned.
5616 The autovectorizer, when vectorizing a load operation from an address
5617 @var{addr} that may be unaligned, will generate two vector loads from
5618 the two aligned addresses around @var{addr}. It then generates a
5619 @code{REALIGN_LOAD} operation to extract the relevant data from the
5620 two loaded vectors. The first two arguments to @code{REALIGN_LOAD},
5621 @var{v1} and @var{v2}, are the two vectors, each of size @var{VS}, and
5622 the third argument, @var{OFF}, defines how the data will be extracted
5623 from these two vectors: if @var{OFF} is 0, then the returned vector is
5624 @var{v2}; otherwise, the returned vector is composed from the last
5625 @var{VS}-@var{OFF} elements of @var{v1} concatenated to the first
5626 @var{OFF} elements of @var{v2}.
5628 If this hook is defined, the autovectorizer will generate a call
5629 to @var{f} (using the DECL tree that this hook returns) and will
5630 use the return value of @var{f} as the argument @var{OFF} to
5631 @code{REALIGN_LOAD}. Therefore, the mask @var{m} returned by @var{f}
5632 should comply with the semantics expected by @code{REALIGN_LOAD}
5634 If this hook is not defined, then @var{addr} will be used as
5635 the argument @var{OFF} to @code{REALIGN_LOAD}, in which case the low
5636 log2(@var{VS}) @minus{} 1 bits of @var{addr} will be considered.
5639 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN (tree @var{x})
5640 This hook should return the DECL of a function @var{f} that implements
5641 widening multiplication of the even elements of two input vectors of type @var{x}.
5643 If this hook is defined, the autovectorizer will use it along with the
5644 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD} target hook when vectorizing
5645 widening multiplication in cases that the order of the results does not have to be
5646 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5647 @code{widen_mult_hi/lo} idioms will be used.
5650 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_ODD (tree @var{x})
5651 This hook should return the DECL of a function @var{f} that implements
5652 widening multiplication of the odd elements of two input vectors of type @var{x}.
5654 If this hook is defined, the autovectorizer will use it along with the
5655 @code{TARGET_VECTORIZE_BUILTIN_MUL_WIDEN_EVEN} target hook when vectorizing
5656 widening multiplication in cases that the order of the results does not have to be
5657 preserved (e.g.@: used only by a reduction computation). Otherwise, the
5658 @code{widen_mult_hi/lo} idioms will be used.
5661 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_CONVERSION (enum tree_code @var{code}, tree @var{type})
5662 This hook should return the DECL of a function that implements conversion of the
5663 input vector of type @var{type}.
5664 If @var{type} is an integral type, the result of the conversion is a vector of
5665 floating-point type of the same size.
5666 If @var{type} is a floating-point type, the result of the conversion is a vector
5667 of integral type of the same size.
5668 @var{code} specifies how the conversion is to be applied
5669 (truncation, rounding, etc.).
5671 If this hook is defined, the autovectorizer will use the
5672 @code{TARGET_VECTORIZE_BUILTIN_CONVERSION} target hook when vectorizing
5673 conversion. Otherwise, it will return @code{NULL_TREE}.
5676 @deftypefn {Target Hook} tree TARGET_VECTORIZE_BUILTIN_VECTORIZED_FUNCTION (enum built_in_function @var{code}, tree @var{vec_type_out}, tree @var{vec_type_in})
5677 This hook should return the decl of a function that implements the vectorized
5678 variant of the builtin function with builtin function code @var{code} or
5679 @code{NULL_TREE} if such a function is not available. The return type of
5680 the vectorized function shall be of vector type @var{vec_type_out} and the
5681 argument types should be @var{vec_type_in}.
5684 @deftypefn {Target Hook} bool TARGET_SUPPORT_VECTOR_MISALIGNMENT (enum machine_mode @var{mode}, tree @var{type}, int @var{misalignment}, bool @var{is_packed})
5685 This hook should return true if the target supports misaligned vector
5686 store/load of a specific factor denoted in the @var{misalignment}
5687 parameter. The vector store/load should be of machine mode @var{mode} and
5688 the elements in the vectors should be of type @var{type}. @var{is_packed}
5689 parameter is true if the memory access is defined in a packed struct.
5692 @node Anchored Addresses
5693 @section Anchored Addresses
5694 @cindex anchored addresses
5695 @cindex @option{-fsection-anchors}
5697 GCC usually addresses every static object as a separate entity.
5698 For example, if we have:
5702 int foo (void) @{ return a + b + c; @}
5705 the code for @code{foo} will usually calculate three separate symbolic
5706 addresses: those of @code{a}, @code{b} and @code{c}. On some targets,
5707 it would be better to calculate just one symbolic address and access
5708 the three variables relative to it. The equivalent pseudocode would
5714 register int *xr = &x;
5715 return xr[&a - &x] + xr[&b - &x] + xr[&c - &x];
5719 (which isn't valid C). We refer to shared addresses like @code{x} as
5720 ``section anchors''. Their use is controlled by @option{-fsection-anchors}.
5722 The hooks below describe the target properties that GCC needs to know
5723 in order to make effective use of section anchors. It won't use
5724 section anchors at all unless either @code{TARGET_MIN_ANCHOR_OFFSET}
5725 or @code{TARGET_MAX_ANCHOR_OFFSET} is set to a nonzero value.
5727 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MIN_ANCHOR_OFFSET
5728 The minimum offset that should be applied to a section anchor.
5729 On most targets, it should be the smallest offset that can be
5730 applied to a base register while still giving a legitimate address
5731 for every mode. The default value is 0.
5734 @deftypevr {Target Hook} HOST_WIDE_INT TARGET_MAX_ANCHOR_OFFSET
5735 Like @code{TARGET_MIN_ANCHOR_OFFSET}, but the maximum (inclusive)
5736 offset that should be applied to section anchors. The default
5740 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_ANCHOR (rtx @var{x})
5741 Write the assembly code to define section anchor @var{x}, which is a
5742 @code{SYMBOL_REF} for which @samp{SYMBOL_REF_ANCHOR_P (@var{x})} is true.
5743 The hook is called with the assembly output position set to the beginning
5744 of @code{SYMBOL_REF_BLOCK (@var{x})}.
5746 If @code{ASM_OUTPUT_DEF} is available, the hook's default definition uses
5747 it to define the symbol as @samp{. + SYMBOL_REF_BLOCK_OFFSET (@var{x})}.
5748 If @code{ASM_OUTPUT_DEF} is not available, the hook's default definition
5749 is @code{NULL}, which disables the use of section anchors altogether.
5752 @deftypefn {Target Hook} bool TARGET_USE_ANCHORS_FOR_SYMBOL_P (rtx @var{x})
5753 Return true if GCC should attempt to use anchors to access @code{SYMBOL_REF}
5754 @var{x}. You can assume @samp{SYMBOL_REF_HAS_BLOCK_INFO_P (@var{x})} and
5755 @samp{!SYMBOL_REF_ANCHOR_P (@var{x})}.
5757 The default version is correct for most targets, but you might need to
5758 intercept this hook to handle things like target-specific attributes
5759 or target-specific sections.
5762 @node Condition Code
5763 @section Condition Code Status
5764 @cindex condition code status
5766 The macros in this section can be split in two families, according to the
5767 two ways of representing condition codes in GCC.
5769 The first representation is the so called @code{(cc0)} representation
5770 (@pxref{Jump Patterns}), where all instructions can have an implicit
5771 clobber of the condition codes. The second is the condition code
5772 register representation, which provides better schedulability for
5773 architectures that do have a condition code register, but on which
5774 most instructions do not affect it. The latter category includes
5777 The implicit clobbering poses a strong restriction on the placement of
5778 the definition and use of the condition code, which need to be in adjacent
5779 insns for machines using @code{(cc0)}. This can prevent important
5780 optimizations on some machines. For example, on the IBM RS/6000, there
5781 is a delay for taken branches unless the condition code register is set
5782 three instructions earlier than the conditional branch. The instruction
5783 scheduler cannot perform this optimization if it is not permitted to
5784 separate the definition and use of the condition code register.
5786 For this reason, it is possible and suggested to use a register to
5787 represent the condition code for new ports. If there is a specific
5788 condition code register in the machine, use a hard register. If the
5789 condition code or comparison result can be placed in any general register,
5790 or if there are multiple condition registers, use a pseudo register.
5791 Registers used to store the condition code value will usually have a mode
5792 that is in class @code{MODE_CC}.
5794 Alternatively, you can use @code{BImode} if the comparison operator is
5795 specified already in the compare instruction. In this case, you are not
5796 interested in most macros in this section.
5799 * CC0 Condition Codes:: Old style representation of condition codes.
5800 * MODE_CC Condition Codes:: Modern representation of condition codes.
5801 * Cond. Exec. Macros:: Macros to control conditional execution.
5804 @node CC0 Condition Codes
5805 @subsection Representation of condition codes using @code{(cc0)}
5809 The file @file{conditions.h} defines a variable @code{cc_status} to
5810 describe how the condition code was computed (in case the interpretation of
5811 the condition code depends on the instruction that it was set by). This
5812 variable contains the RTL expressions on which the condition code is
5813 currently based, and several standard flags.
5815 Sometimes additional machine-specific flags must be defined in the machine
5816 description header file. It can also add additional machine-specific
5817 information by defining @code{CC_STATUS_MDEP}.
5819 @defmac CC_STATUS_MDEP
5820 C code for a data type which is used for declaring the @code{mdep}
5821 component of @code{cc_status}. It defaults to @code{int}.
5823 This macro is not used on machines that do not use @code{cc0}.
5826 @defmac CC_STATUS_MDEP_INIT
5827 A C expression to initialize the @code{mdep} field to ``empty''.
5828 The default definition does nothing, since most machines don't use
5829 the field anyway. If you want to use the field, you should probably
5830 define this macro to initialize it.
5832 This macro is not used on machines that do not use @code{cc0}.
5835 @defmac NOTICE_UPDATE_CC (@var{exp}, @var{insn})
5836 A C compound statement to set the components of @code{cc_status}
5837 appropriately for an insn @var{insn} whose body is @var{exp}. It is
5838 this macro's responsibility to recognize insns that set the condition
5839 code as a byproduct of other activity as well as those that explicitly
5842 This macro is not used on machines that do not use @code{cc0}.
5844 If there are insns that do not set the condition code but do alter
5845 other machine registers, this macro must check to see whether they
5846 invalidate the expressions that the condition code is recorded as
5847 reflecting. For example, on the 68000, insns that store in address
5848 registers do not set the condition code, which means that usually
5849 @code{NOTICE_UPDATE_CC} can leave @code{cc_status} unaltered for such
5850 insns. But suppose that the previous insn set the condition code
5851 based on location @samp{a4@@(102)} and the current insn stores a new
5852 value in @samp{a4}. Although the condition code is not changed by
5853 this, it will no longer be true that it reflects the contents of
5854 @samp{a4@@(102)}. Therefore, @code{NOTICE_UPDATE_CC} must alter
5855 @code{cc_status} in this case to say that nothing is known about the
5856 condition code value.
5858 The definition of @code{NOTICE_UPDATE_CC} must be prepared to deal
5859 with the results of peephole optimization: insns whose patterns are
5860 @code{parallel} RTXs containing various @code{reg}, @code{mem} or
5861 constants which are just the operands. The RTL structure of these
5862 insns is not sufficient to indicate what the insns actually do. What
5863 @code{NOTICE_UPDATE_CC} should do when it sees one is just to run
5864 @code{CC_STATUS_INIT}.
5866 A possible definition of @code{NOTICE_UPDATE_CC} is to call a function
5867 that looks at an attribute (@pxref{Insn Attributes}) named, for example,
5868 @samp{cc}. This avoids having detailed information about patterns in
5869 two places, the @file{md} file and in @code{NOTICE_UPDATE_CC}.
5872 @node MODE_CC Condition Codes
5873 @subsection Representation of condition codes using registers
5877 @defmac SELECT_CC_MODE (@var{op}, @var{x}, @var{y})
5878 On many machines, the condition code may be produced by other instructions
5879 than compares, for example the branch can use directly the condition
5880 code set by a subtract instruction. However, on some machines
5881 when the condition code is set this way some bits (such as the overflow
5882 bit) are not set in the same way as a test instruction, so that a different
5883 branch instruction must be used for some conditional branches. When
5884 this happens, use the machine mode of the condition code register to
5885 record different formats of the condition code register. Modes can
5886 also be used to record which compare instruction (e.g. a signed or an
5887 unsigned comparison) produced the condition codes.
5889 If other modes than @code{CCmode} are required, add them to
5890 @file{@var{machine}-modes.def} and define @code{SELECT_CC_MODE} to choose
5891 a mode given an operand of a compare. This is needed because the modes
5892 have to be chosen not only during RTL generation but also, for example,
5893 by instruction combination. The result of @code{SELECT_CC_MODE} should
5894 be consistent with the mode used in the patterns; for example to support
5895 the case of the add on the SPARC discussed above, we have the pattern
5899 [(set (reg:CC_NOOV 0)
5901 (plus:SI (match_operand:SI 0 "register_operand" "%r")
5902 (match_operand:SI 1 "arith_operand" "rI"))
5909 together with a @code{SELECT_CC_MODE} that returns @code{CC_NOOVmode}
5910 for comparisons whose argument is a @code{plus}:
5913 #define SELECT_CC_MODE(OP,X,Y) \
5914 (GET_MODE_CLASS (GET_MODE (X)) == MODE_FLOAT \
5915 ? ((OP == EQ || OP == NE) ? CCFPmode : CCFPEmode) \
5916 : ((GET_CODE (X) == PLUS || GET_CODE (X) == MINUS \
5917 || GET_CODE (X) == NEG) \
5918 ? CC_NOOVmode : CCmode))
5921 Another reason to use modes is to retain information on which operands
5922 were used by the comparison; see @code{REVERSIBLE_CC_MODE} later in
5925 You should define this macro if and only if you define extra CC modes
5926 in @file{@var{machine}-modes.def}.
5929 @defmac CANONICALIZE_COMPARISON (@var{code}, @var{op0}, @var{op1})
5930 On some machines not all possible comparisons are defined, but you can
5931 convert an invalid comparison into a valid one. For example, the Alpha
5932 does not have a @code{GT} comparison, but you can use an @code{LT}
5933 comparison instead and swap the order of the operands.
5935 On such machines, define this macro to be a C statement to do any
5936 required conversions. @var{code} is the initial comparison code
5937 and @var{op0} and @var{op1} are the left and right operands of the
5938 comparison, respectively. You should modify @var{code}, @var{op0}, and
5939 @var{op1} as required.
5941 GCC will not assume that the comparison resulting from this macro is
5942 valid but will see if the resulting insn matches a pattern in the
5945 You need not define this macro if it would never change the comparison
5949 @defmac REVERSIBLE_CC_MODE (@var{mode})
5950 A C expression whose value is one if it is always safe to reverse a
5951 comparison whose mode is @var{mode}. If @code{SELECT_CC_MODE}
5952 can ever return @var{mode} for a floating-point inequality comparison,
5953 then @code{REVERSIBLE_CC_MODE (@var{mode})} must be zero.
5955 You need not define this macro if it would always returns zero or if the
5956 floating-point format is anything other than @code{IEEE_FLOAT_FORMAT}.
5957 For example, here is the definition used on the SPARC, where floating-point
5958 inequality comparisons are always given @code{CCFPEmode}:
5961 #define REVERSIBLE_CC_MODE(MODE) ((MODE) != CCFPEmode)
5965 @defmac REVERSE_CONDITION (@var{code}, @var{mode})
5966 A C expression whose value is reversed condition code of the @var{code} for
5967 comparison done in CC_MODE @var{mode}. The macro is used only in case
5968 @code{REVERSIBLE_CC_MODE (@var{mode})} is nonzero. Define this macro in case
5969 machine has some non-standard way how to reverse certain conditionals. For
5970 instance in case all floating point conditions are non-trapping, compiler may
5971 freely convert unordered compares to ordered one. Then definition may look
5975 #define REVERSE_CONDITION(CODE, MODE) \
5976 ((MODE) != CCFPmode ? reverse_condition (CODE) \
5977 : reverse_condition_maybe_unordered (CODE))
5981 @deftypefn {Target Hook} bool TARGET_FIXED_CONDITION_CODE_REGS (unsigned int *, unsigned int *)
5982 On targets which do not use @code{(cc0)}, and which use a hard
5983 register rather than a pseudo-register to hold condition codes, the
5984 regular CSE passes are often not able to identify cases in which the
5985 hard register is set to a common value. Use this hook to enable a
5986 small pass which optimizes such cases. This hook should return true
5987 to enable this pass, and it should set the integers to which its
5988 arguments point to the hard register numbers used for condition codes.
5989 When there is only one such register, as is true on most systems, the
5990 integer pointed to by the second argument should be set to
5991 @code{INVALID_REGNUM}.
5993 The default version of this hook returns false.
5996 @deftypefn {Target Hook} enum machine_mode TARGET_CC_MODES_COMPATIBLE (enum machine_mode, enum machine_mode)
5997 On targets which use multiple condition code modes in class
5998 @code{MODE_CC}, it is sometimes the case that a comparison can be
5999 validly done in more than one mode. On such a system, define this
6000 target hook to take two mode arguments and to return a mode in which
6001 both comparisons may be validly done. If there is no such mode,
6002 return @code{VOIDmode}.
6004 The default version of this hook checks whether the modes are the
6005 same. If they are, it returns that mode. If they are different, it
6006 returns @code{VOIDmode}.
6009 @node Cond. Exec. Macros
6010 @subsection Macros to control conditional execution
6011 @findex conditional execution
6014 There is one macro that may need to be defined for targets
6015 supporting conditional execution, independent of how they
6016 represent conditional branches.
6018 @defmac REVERSE_CONDEXEC_PREDICATES_P (@var{op1}, @var{op2})
6019 A C expression that returns true if the conditional execution predicate
6020 @var{op1}, a comparison operation, is the inverse of @var{op2} and vice
6021 versa. Define this to return 0 if the target has conditional execution
6022 predicates that cannot be reversed safely. There is no need to validate
6023 that the arguments of op1 and op2 are the same, this is done separately.
6024 If no expansion is specified, this macro is defined as follows:
6027 #define REVERSE_CONDEXEC_PREDICATES_P (x, y) \
6028 (GET_CODE ((x)) == reversed_comparison_code ((y), NULL))
6033 @section Describing Relative Costs of Operations
6034 @cindex costs of instructions
6035 @cindex relative costs
6036 @cindex speed of instructions
6038 These macros let you describe the relative speed of various operations
6039 on the target machine.
6041 @defmac REGISTER_MOVE_COST (@var{mode}, @var{from}, @var{to})
6042 A C expression for the cost of moving data of mode @var{mode} from a
6043 register in class @var{from} to one in class @var{to}. The classes are
6044 expressed using the enumeration values such as @code{GENERAL_REGS}. A
6045 value of 2 is the default; other values are interpreted relative to
6048 It is not required that the cost always equal 2 when @var{from} is the
6049 same as @var{to}; on some machines it is expensive to move between
6050 registers if they are not general registers.
6052 If reload sees an insn consisting of a single @code{set} between two
6053 hard registers, and if @code{REGISTER_MOVE_COST} applied to their
6054 classes returns a value of 2, reload does not check to ensure that the
6055 constraints of the insn are met. Setting a cost of other than 2 will
6056 allow reload to verify that the constraints are met. You should do this
6057 if the @samp{mov@var{m}} pattern's constraints do not allow such copying.
6060 @defmac MEMORY_MOVE_COST (@var{mode}, @var{class}, @var{in})
6061 A C expression for the cost of moving data of mode @var{mode} between a
6062 register of class @var{class} and memory; @var{in} is zero if the value
6063 is to be written to memory, nonzero if it is to be read in. This cost
6064 is relative to those in @code{REGISTER_MOVE_COST}. If moving between
6065 registers and memory is more expensive than between two registers, you
6066 should define this macro to express the relative cost.
6068 If you do not define this macro, GCC uses a default cost of 4 plus
6069 the cost of copying via a secondary reload register, if one is
6070 needed. If your machine requires a secondary reload register to copy
6071 between memory and a register of @var{class} but the reload mechanism is
6072 more complex than copying via an intermediate, define this macro to
6073 reflect the actual cost of the move.
6075 GCC defines the function @code{memory_move_secondary_cost} if
6076 secondary reloads are needed. It computes the costs due to copying via
6077 a secondary register. If your machine copies from memory using a
6078 secondary register in the conventional way but the default base value of
6079 4 is not correct for your machine, define this macro to add some other
6080 value to the result of that function. The arguments to that function
6081 are the same as to this macro.
6084 @defmac BRANCH_COST (@var{speed_p}, @var{predictable_p})
6085 A C expression for the cost of a branch instruction. A value of 1 is the
6086 default; other values are interpreted relative to that. Parameter @var{speed_p}
6087 is true when the branch in question should be optimized for speed. When
6088 it is false, @code{BRANCH_COST} should be returning value optimal for code size
6089 rather then performance considerations. @var{predictable_p} is true for well
6090 predictable branches. On many architectures the @code{BRANCH_COST} can be
6094 Here are additional macros which do not specify precise relative costs,
6095 but only that certain actions are more expensive than GCC would
6098 @defmac SLOW_BYTE_ACCESS
6099 Define this macro as a C expression which is nonzero if accessing less
6100 than a word of memory (i.e.@: a @code{char} or a @code{short}) is no
6101 faster than accessing a word of memory, i.e., if such access
6102 require more than one instruction or if there is no difference in cost
6103 between byte and (aligned) word loads.
6105 When this macro is not defined, the compiler will access a field by
6106 finding the smallest containing object; when it is defined, a fullword
6107 load will be used if alignment permits. Unless bytes accesses are
6108 faster than word accesses, using word accesses is preferable since it
6109 may eliminate subsequent memory access if subsequent accesses occur to
6110 other fields in the same word of the structure, but to different bytes.
6113 @defmac SLOW_UNALIGNED_ACCESS (@var{mode}, @var{alignment})
6114 Define this macro to be the value 1 if memory accesses described by the
6115 @var{mode} and @var{alignment} parameters have a cost many times greater
6116 than aligned accesses, for example if they are emulated in a trap
6119 When this macro is nonzero, the compiler will act as if
6120 @code{STRICT_ALIGNMENT} were nonzero when generating code for block
6121 moves. This can cause significantly more instructions to be produced.
6122 Therefore, do not set this macro nonzero if unaligned accesses only add a
6123 cycle or two to the time for a memory access.
6125 If the value of this macro is always zero, it need not be defined. If
6126 this macro is defined, it should produce a nonzero value when
6127 @code{STRICT_ALIGNMENT} is nonzero.
6131 The threshold of number of scalar memory-to-memory move insns, @emph{below}
6132 which a sequence of insns should be generated instead of a
6133 string move insn or a library call. Increasing the value will always
6134 make code faster, but eventually incurs high cost in increased code size.
6136 Note that on machines where the corresponding move insn is a
6137 @code{define_expand} that emits a sequence of insns, this macro counts
6138 the number of such sequences.
6140 If you don't define this, a reasonable default is used.
6143 @defmac MOVE_BY_PIECES_P (@var{size}, @var{alignment})
6144 A C expression used to determine whether @code{move_by_pieces} will be used to
6145 copy a chunk of memory, or whether some other block move mechanism
6146 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6147 than @code{MOVE_RATIO}.
6150 @defmac MOVE_MAX_PIECES
6151 A C expression used by @code{move_by_pieces} to determine the largest unit
6152 a load or store used to copy memory is. Defaults to @code{MOVE_MAX}.
6156 The threshold of number of scalar move insns, @emph{below} which a sequence
6157 of insns should be generated to clear memory instead of a string clear insn
6158 or a library call. Increasing the value will always make code faster, but
6159 eventually incurs high cost in increased code size.
6161 If you don't define this, a reasonable default is used.
6164 @defmac CLEAR_BY_PIECES_P (@var{size}, @var{alignment})
6165 A C expression used to determine whether @code{clear_by_pieces} will be used
6166 to clear a chunk of memory, or whether some other block clear mechanism
6167 will be used. Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6168 than @code{CLEAR_RATIO}.
6172 The threshold of number of scalar move insns, @emph{below} which a sequence
6173 of insns should be generated to set memory to a constant value, instead of
6174 a block set insn or a library call.
6175 Increasing the value will always make code faster, but
6176 eventually incurs high cost in increased code size.
6178 If you don't define this, it defaults to the value of @code{MOVE_RATIO}.
6181 @defmac SET_BY_PIECES_P (@var{size}, @var{alignment})
6182 A C expression used to determine whether @code{store_by_pieces} will be
6183 used to set a chunk of memory to a constant value, or whether some
6184 other mechanism will be used. Used by @code{__builtin_memset} when
6185 storing values other than constant zero.
6186 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6187 than @code{SET_RATIO}.
6190 @defmac STORE_BY_PIECES_P (@var{size}, @var{alignment})
6191 A C expression used to determine whether @code{store_by_pieces} will be
6192 used to set a chunk of memory to a constant string value, or whether some
6193 other mechanism will be used. Used by @code{__builtin_strcpy} when
6194 called with a constant source string.
6195 Defaults to 1 if @code{move_by_pieces_ninsns} returns less
6196 than @code{MOVE_RATIO}.
6199 @defmac USE_LOAD_POST_INCREMENT (@var{mode})
6200 A C expression used to determine whether a load postincrement is a good
6201 thing to use for a given mode. Defaults to the value of
6202 @code{HAVE_POST_INCREMENT}.
6205 @defmac USE_LOAD_POST_DECREMENT (@var{mode})
6206 A C expression used to determine whether a load postdecrement is a good
6207 thing to use for a given mode. Defaults to the value of
6208 @code{HAVE_POST_DECREMENT}.
6211 @defmac USE_LOAD_PRE_INCREMENT (@var{mode})
6212 A C expression used to determine whether a load preincrement is a good
6213 thing to use for a given mode. Defaults to the value of
6214 @code{HAVE_PRE_INCREMENT}.
6217 @defmac USE_LOAD_PRE_DECREMENT (@var{mode})
6218 A C expression used to determine whether a load predecrement is a good
6219 thing to use for a given mode. Defaults to the value of
6220 @code{HAVE_PRE_DECREMENT}.
6223 @defmac USE_STORE_POST_INCREMENT (@var{mode})
6224 A C expression used to determine whether a store postincrement is a good
6225 thing to use for a given mode. Defaults to the value of
6226 @code{HAVE_POST_INCREMENT}.
6229 @defmac USE_STORE_POST_DECREMENT (@var{mode})
6230 A C expression used to determine whether a store postdecrement is a good
6231 thing to use for a given mode. Defaults to the value of
6232 @code{HAVE_POST_DECREMENT}.
6235 @defmac USE_STORE_PRE_INCREMENT (@var{mode})
6236 This macro is used to determine whether a store preincrement is a good
6237 thing to use for a given mode. Defaults to the value of
6238 @code{HAVE_PRE_INCREMENT}.
6241 @defmac USE_STORE_PRE_DECREMENT (@var{mode})
6242 This macro is used to determine whether a store predecrement is a good
6243 thing to use for a given mode. Defaults to the value of
6244 @code{HAVE_PRE_DECREMENT}.
6247 @defmac NO_FUNCTION_CSE
6248 Define this macro if it is as good or better to call a constant
6249 function address than to call an address kept in a register.
6252 @defmac RANGE_TEST_NON_SHORT_CIRCUIT
6253 Define this macro if a non-short-circuit operation produced by
6254 @samp{fold_range_test ()} is optimal. This macro defaults to true if
6255 @code{BRANCH_COST} is greater than or equal to the value 2.
6258 @deftypefn {Target Hook} bool TARGET_RTX_COSTS (rtx @var{x}, int @var{code}, int @var{outer_code}, int *@var{total})
6259 This target hook describes the relative costs of RTL expressions.
6261 The cost may depend on the precise form of the expression, which is
6262 available for examination in @var{x}, and the rtx code of the expression
6263 in which it is contained, found in @var{outer_code}. @var{code} is the
6264 expression code---redundant, since it can be obtained with
6265 @code{GET_CODE (@var{x})}.
6267 In implementing this hook, you can use the construct
6268 @code{COSTS_N_INSNS (@var{n})} to specify a cost equal to @var{n} fast
6271 On entry to the hook, @code{*@var{total}} contains a default estimate
6272 for the cost of the expression. The hook should modify this value as
6273 necessary. Traditionally, the default costs are @code{COSTS_N_INSNS (5)}
6274 for multiplications, @code{COSTS_N_INSNS (7)} for division and modulus
6275 operations, and @code{COSTS_N_INSNS (1)} for all other operations.
6277 When optimizing for code size, i.e.@: when @code{optimize_size} is
6278 nonzero, this target hook should be used to estimate the relative
6279 size cost of an expression, again relative to @code{COSTS_N_INSNS}.
6281 The hook returns true when all subexpressions of @var{x} have been
6282 processed, and false when @code{rtx_cost} should recurse.
6285 @deftypefn {Target Hook} int TARGET_ADDRESS_COST (rtx @var{address})
6286 This hook computes the cost of an addressing mode that contains
6287 @var{address}. If not defined, the cost is computed from
6288 the @var{address} expression and the @code{TARGET_RTX_COST} hook.
6290 For most CISC machines, the default cost is a good approximation of the
6291 true cost of the addressing mode. However, on RISC machines, all
6292 instructions normally have the same length and execution time. Hence
6293 all addresses will have equal costs.
6295 In cases where more than one form of an address is known, the form with
6296 the lowest cost will be used. If multiple forms have the same, lowest,
6297 cost, the one that is the most complex will be used.
6299 For example, suppose an address that is equal to the sum of a register
6300 and a constant is used twice in the same basic block. When this macro
6301 is not defined, the address will be computed in a register and memory
6302 references will be indirect through that register. On machines where
6303 the cost of the addressing mode containing the sum is no higher than
6304 that of a simple indirect reference, this will produce an additional
6305 instruction and possibly require an additional register. Proper
6306 specification of this macro eliminates this overhead for such machines.
6308 This hook is never called with an invalid address.
6310 On machines where an address involving more than one register is as
6311 cheap as an address computation involving only one register, defining
6312 @code{TARGET_ADDRESS_COST} to reflect this can cause two registers to
6313 be live over a region of code where only one would have been if
6314 @code{TARGET_ADDRESS_COST} were not defined in that manner. This effect
6315 should be considered in the definition of this macro. Equivalent costs
6316 should probably only be given to addresses with different numbers of
6317 registers on machines with lots of registers.
6321 @section Adjusting the Instruction Scheduler
6323 The instruction scheduler may need a fair amount of machine-specific
6324 adjustment in order to produce good code. GCC provides several target
6325 hooks for this purpose. It is usually enough to define just a few of
6326 them: try the first ones in this list first.
6328 @deftypefn {Target Hook} int TARGET_SCHED_ISSUE_RATE (void)
6329 This hook returns the maximum number of instructions that can ever
6330 issue at the same time on the target machine. The default is one.
6331 Although the insn scheduler can define itself the possibility of issue
6332 an insn on the same cycle, the value can serve as an additional
6333 constraint to issue insns on the same simulated processor cycle (see
6334 hooks @samp{TARGET_SCHED_REORDER} and @samp{TARGET_SCHED_REORDER2}).
6335 This value must be constant over the entire compilation. If you need
6336 it to vary depending on what the instructions are, you must use
6337 @samp{TARGET_SCHED_VARIABLE_ISSUE}.
6340 @deftypefn {Target Hook} int TARGET_SCHED_VARIABLE_ISSUE (FILE *@var{file}, int @var{verbose}, rtx @var{insn}, int @var{more})
6341 This hook is executed by the scheduler after it has scheduled an insn
6342 from the ready list. It should return the number of insns which can
6343 still be issued in the current cycle. The default is
6344 @samp{@w{@var{more} - 1}} for insns other than @code{CLOBBER} and
6345 @code{USE}, which normally are not counted against the issue rate.
6346 You should define this hook if some insns take more machine resources
6347 than others, so that fewer insns can follow them in the same cycle.
6348 @var{file} is either a null pointer, or a stdio stream to write any
6349 debug output to. @var{verbose} is the verbose level provided by
6350 @option{-fsched-verbose-@var{n}}. @var{insn} is the instruction that
6354 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_COST (rtx @var{insn}, rtx @var{link}, rtx @var{dep_insn}, int @var{cost})
6355 This function corrects the value of @var{cost} based on the
6356 relationship between @var{insn} and @var{dep_insn} through the
6357 dependence @var{link}. It should return the new value. The default
6358 is to make no adjustment to @var{cost}. This can be used for example
6359 to specify to the scheduler using the traditional pipeline description
6360 that an output- or anti-dependence does not incur the same cost as a
6361 data-dependence. If the scheduler using the automaton based pipeline
6362 description, the cost of anti-dependence is zero and the cost of
6363 output-dependence is maximum of one and the difference of latency
6364 times of the first and the second insns. If these values are not
6365 acceptable, you could use the hook to modify them too. See also
6366 @pxref{Processor pipeline description}.
6369 @deftypefn {Target Hook} int TARGET_SCHED_ADJUST_PRIORITY (rtx @var{insn}, int @var{priority})
6370 This hook adjusts the integer scheduling priority @var{priority} of
6371 @var{insn}. It should return the new priority. Increase the priority to
6372 execute @var{insn} earlier, reduce the priority to execute @var{insn}
6373 later. Do not define this hook if you do not need to adjust the
6374 scheduling priorities of insns.
6377 @deftypefn {Target Hook} int TARGET_SCHED_REORDER (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_readyp}, int @var{clock})
6378 This hook is executed by the scheduler after it has scheduled the ready
6379 list, to allow the machine description to reorder it (for example to
6380 combine two small instructions together on @samp{VLIW} machines).
6381 @var{file} is either a null pointer, or a stdio stream to write any
6382 debug output to. @var{verbose} is the verbose level provided by
6383 @option{-fsched-verbose-@var{n}}. @var{ready} is a pointer to the ready
6384 list of instructions that are ready to be scheduled. @var{n_readyp} is
6385 a pointer to the number of elements in the ready list. The scheduler
6386 reads the ready list in reverse order, starting with
6387 @var{ready}[@var{*n_readyp}-1] and going to @var{ready}[0]. @var{clock}
6388 is the timer tick of the scheduler. You may modify the ready list and
6389 the number of ready insns. The return value is the number of insns that
6390 can issue this cycle; normally this is just @code{issue_rate}. See also
6391 @samp{TARGET_SCHED_REORDER2}.
6394 @deftypefn {Target Hook} int TARGET_SCHED_REORDER2 (FILE *@var{file}, int @var{verbose}, rtx *@var{ready}, int *@var{n_ready}, @var{clock})
6395 Like @samp{TARGET_SCHED_REORDER}, but called at a different time. That
6396 function is called whenever the scheduler starts a new cycle. This one
6397 is called once per iteration over a cycle, immediately after
6398 @samp{TARGET_SCHED_VARIABLE_ISSUE}; it can reorder the ready list and
6399 return the number of insns to be scheduled in the same cycle. Defining
6400 this hook can be useful if there are frequent situations where
6401 scheduling one insn causes other insns to become ready in the same
6402 cycle. These other insns can then be taken into account properly.
6405 @deftypefn {Target Hook} void TARGET_SCHED_DEPENDENCIES_EVALUATION_HOOK (rtx @var{head}, rtx @var{tail})
6406 This hook is called after evaluation forward dependencies of insns in
6407 chain given by two parameter values (@var{head} and @var{tail}
6408 correspondingly) but before insns scheduling of the insn chain. For
6409 example, it can be used for better insn classification if it requires
6410 analysis of dependencies. This hook can use backward and forward
6411 dependencies of the insn scheduler because they are already
6415 @deftypefn {Target Hook} void TARGET_SCHED_INIT (FILE *@var{file}, int @var{verbose}, int @var{max_ready})
6416 This hook is executed by the scheduler at the beginning of each block of
6417 instructions that are to be scheduled. @var{file} is either a null
6418 pointer, or a stdio stream to write any debug output to. @var{verbose}
6419 is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6420 @var{max_ready} is the maximum number of insns in the current scheduling
6421 region that can be live at the same time. This can be used to allocate
6422 scratch space if it is needed, e.g.@: by @samp{TARGET_SCHED_REORDER}.
6425 @deftypefn {Target Hook} void TARGET_SCHED_FINISH (FILE *@var{file}, int @var{verbose})
6426 This hook is executed by the scheduler at the end of each block of
6427 instructions that are to be scheduled. It can be used to perform
6428 cleanup of any actions done by the other scheduling hooks. @var{file}
6429 is either a null pointer, or a stdio stream to write any debug output
6430 to. @var{verbose} is the verbose level provided by
6431 @option{-fsched-verbose-@var{n}}.
6434 @deftypefn {Target Hook} void TARGET_SCHED_INIT_GLOBAL (FILE *@var{file}, int @var{verbose}, int @var{old_max_uid})
6435 This hook is executed by the scheduler after function level initializations.
6436 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6437 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6438 @var{old_max_uid} is the maximum insn uid when scheduling begins.
6441 @deftypefn {Target Hook} void TARGET_SCHED_FINISH_GLOBAL (FILE *@var{file}, int @var{verbose})
6442 This is the cleanup hook corresponding to @code{TARGET_SCHED_INIT_GLOBAL}.
6443 @var{file} is either a null pointer, or a stdio stream to write any debug output to.
6444 @var{verbose} is the verbose level provided by @option{-fsched-verbose-@var{n}}.
6447 @deftypefn {Target Hook} int TARGET_SCHED_DFA_PRE_CYCLE_INSN (void)
6448 The hook returns an RTL insn. The automaton state used in the
6449 pipeline hazard recognizer is changed as if the insn were scheduled
6450 when the new simulated processor cycle starts. Usage of the hook may
6451 simplify the automaton pipeline description for some @acronym{VLIW}
6452 processors. If the hook is defined, it is used only for the automaton
6453 based pipeline description. The default is not to change the state
6454 when the new simulated processor cycle starts.
6457 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN (void)
6458 The hook can be used to initialize data used by the previous hook.
6461 @deftypefn {Target Hook} int TARGET_SCHED_DFA_POST_CYCLE_INSN (void)
6462 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6463 to changed the state as if the insn were scheduled when the new
6464 simulated processor cycle finishes.
6467 @deftypefn {Target Hook} void TARGET_SCHED_INIT_DFA_POST_CYCLE_INSN (void)
6468 The hook is analogous to @samp{TARGET_SCHED_INIT_DFA_PRE_CYCLE_INSN} but
6469 used to initialize data used by the previous hook.
6472 @deftypefn {Target Hook} void TARGET_SCHED_DFA_PRE_CYCLE_ADVANCE (void)
6473 The hook to notify target that the current simulated cycle is about to finish.
6474 The hook is analogous to @samp{TARGET_SCHED_DFA_PRE_CYCLE_INSN} but used
6475 to change the state in more complicated situations - e.g., when advancing
6476 state on a single insn is not enough.
6479 @deftypefn {Target Hook} void TARGET_SCHED_DFA_POST_CYCLE_ADVANCE (void)
6480 The hook to notify target that new simulated cycle has just started.
6481 The hook is analogous to @samp{TARGET_SCHED_DFA_POST_CYCLE_INSN} but used
6482 to change the state in more complicated situations - e.g., when advancing
6483 state on a single insn is not enough.
6486 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD (void)
6487 This hook controls better choosing an insn from the ready insn queue
6488 for the @acronym{DFA}-based insn scheduler. Usually the scheduler
6489 chooses the first insn from the queue. If the hook returns a positive
6490 value, an additional scheduler code tries all permutations of
6491 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD ()}
6492 subsequent ready insns to choose an insn whose issue will result in
6493 maximal number of issued insns on the same cycle. For the
6494 @acronym{VLIW} processor, the code could actually solve the problem of
6495 packing simple insns into the @acronym{VLIW} insn. Of course, if the
6496 rules of @acronym{VLIW} packing are described in the automaton.
6498 This code also could be used for superscalar @acronym{RISC}
6499 processors. Let us consider a superscalar @acronym{RISC} processor
6500 with 3 pipelines. Some insns can be executed in pipelines @var{A} or
6501 @var{B}, some insns can be executed only in pipelines @var{B} or
6502 @var{C}, and one insn can be executed in pipeline @var{B}. The
6503 processor may issue the 1st insn into @var{A} and the 2nd one into
6504 @var{B}. In this case, the 3rd insn will wait for freeing @var{B}
6505 until the next cycle. If the scheduler issues the 3rd insn the first,
6506 the processor could issue all 3 insns per cycle.
6508 Actually this code demonstrates advantages of the automaton based
6509 pipeline hazard recognizer. We try quickly and easy many insn
6510 schedules to choose the best one.
6512 The default is no multipass scheduling.
6515 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD (rtx)
6517 This hook controls what insns from the ready insn queue will be
6518 considered for the multipass insn scheduling. If the hook returns
6519 zero for insn passed as the parameter, the insn will be not chosen to
6522 The default is that any ready insns can be chosen to be issued.
6525 @deftypefn {Target Hook} int TARGET_SCHED_DFA_NEW_CYCLE (FILE *, int, rtx, int, int, int *)
6527 This hook is called by the insn scheduler before issuing insn passed
6528 as the third parameter on given cycle. If the hook returns nonzero,
6529 the insn is not issued on given processors cycle. Instead of that,
6530 the processor cycle is advanced. If the value passed through the last
6531 parameter is zero, the insn ready queue is not sorted on the new cycle
6532 start as usually. The first parameter passes file for debugging
6533 output. The second one passes the scheduler verbose level of the
6534 debugging output. The forth and the fifth parameter values are
6535 correspondingly processor cycle on which the previous insn has been
6536 issued and the current processor cycle.
6539 @deftypefn {Target Hook} bool TARGET_SCHED_IS_COSTLY_DEPENDENCE (struct dep_def *@var{_dep}, int @var{cost}, int @var{distance})
6540 This hook is used to define which dependences are considered costly by
6541 the target, so costly that it is not advisable to schedule the insns that
6542 are involved in the dependence too close to one another. The parameters
6543 to this hook are as follows: The first parameter @var{_dep} is the dependence
6544 being evaluated. The second parameter @var{cost} is the cost of the
6545 dependence, and the third
6546 parameter @var{distance} is the distance in cycles between the two insns.
6547 The hook returns @code{true} if considering the distance between the two
6548 insns the dependence between them is considered costly by the target,
6549 and @code{false} otherwise.
6551 Defining this hook can be useful in multiple-issue out-of-order machines,
6552 where (a) it's practically hopeless to predict the actual data/resource
6553 delays, however: (b) there's a better chance to predict the actual grouping
6554 that will be formed, and (c) correctly emulating the grouping can be very
6555 important. In such targets one may want to allow issuing dependent insns
6556 closer to one another---i.e., closer than the dependence distance; however,
6557 not in cases of "costly dependences", which this hooks allows to define.
6560 @deftypefn {Target Hook} void TARGET_SCHED_H_I_D_EXTENDED (void)
6561 This hook is called by the insn scheduler after emitting a new instruction to
6562 the instruction stream. The hook notifies a target backend to extend its
6563 per instruction data structures.
6566 @deftypefn {Target Hook} void * TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6567 Return a pointer to a store large enough to hold target scheduling context.
6570 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6571 Initialize store pointed to by @var{tc} to hold target scheduling context.
6572 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6573 beginning of the block. Otherwise, make a copy of the current context in
6577 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6578 Copy target scheduling context pointer to by @var{tc} to the current context.
6581 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6582 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6585 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6586 Deallocate a store for target scheduling context pointed to by @var{tc}.
6589 @deftypefn {Target Hook} void * TARGET_SCHED_ALLOC_SCHED_CONTEXT (void)
6590 Return a pointer to a store large enough to hold target scheduling context.
6593 @deftypefn {Target Hook} void TARGET_SCHED_INIT_SCHED_CONTEXT (void *@var{tc}, bool @var{clean_p})
6594 Initialize store pointed to by @var{tc} to hold target scheduling context.
6595 It @var{clean_p} is true then initialize @var{tc} as if scheduler is at the
6596 beginning of the block. Otherwise, make a copy of the current context in
6600 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_CONTEXT (void *@var{tc})
6601 Copy target scheduling context pointer to by @var{tc} to the current context.
6604 @deftypefn {Target Hook} void TARGET_SCHED_CLEAR_SCHED_CONTEXT (void *@var{tc})
6605 Deallocate internal data in target scheduling context pointed to by @var{tc}.
6608 @deftypefn {Target Hook} void TARGET_SCHED_FREE_SCHED_CONTEXT (void *@var{tc})
6609 Deallocate a store for target scheduling context pointed to by @var{tc}.
6612 @deftypefn {Target Hook} int TARGET_SCHED_SPECULATE_INSN (rtx @var{insn}, int @var{request}, rtx *@var{new_pat})
6613 This hook is called by the insn scheduler when @var{insn} has only
6614 speculative dependencies and therefore can be scheduled speculatively.
6615 The hook is used to check if the pattern of @var{insn} has a speculative
6616 version and, in case of successful check, to generate that speculative
6617 pattern. The hook should return 1, if the instruction has a speculative form,
6618 or @minus{}1, if it doesn't. @var{request} describes the type of requested
6619 speculation. If the return value equals 1 then @var{new_pat} is assigned
6620 the generated speculative pattern.
6623 @deftypefn {Target Hook} int TARGET_SCHED_NEEDS_BLOCK_P (rtx @var{insn})
6624 This hook is called by the insn scheduler during generation of recovery code
6625 for @var{insn}. It should return nonzero, if the corresponding check
6626 instruction should branch to recovery code, or zero otherwise.
6629 @deftypefn {Target Hook} rtx TARGET_SCHED_GEN_CHECK (rtx @var{insn}, rtx @var{label}, int @var{mutate_p})
6630 This hook is called by the insn scheduler to generate a pattern for recovery
6631 check instruction. If @var{mutate_p} is zero, then @var{insn} is a
6632 speculative instruction for which the check should be generated.
6633 @var{label} is either a label of a basic block, where recovery code should
6634 be emitted, or a null pointer, when requested check doesn't branch to
6635 recovery code (a simple check). If @var{mutate_p} is nonzero, then
6636 a pattern for a branchy check corresponding to a simple check denoted by
6637 @var{insn} should be generated. In this case @var{label} can't be null.
6640 @deftypefn {Target Hook} int TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD_SPEC (rtx @var{insn})
6641 This hook is used as a workaround for
6642 @samp{TARGET_SCHED_FIRST_CYCLE_MULTIPASS_DFA_LOOKAHEAD_GUARD} not being
6643 called on the first instruction of the ready list. The hook is used to
6644 discard speculative instruction that stand first in the ready list from
6645 being scheduled on the current cycle. For non-speculative instructions,
6646 the hook should always return nonzero. For example, in the ia64 backend
6647 the hook is used to cancel data speculative insns when the ALAT table
6651 @deftypefn {Target Hook} void TARGET_SCHED_SET_SCHED_FLAGS (unsigned int *@var{flags}, spec_info_t @var{spec_info})
6652 This hook is used by the insn scheduler to find out what features should be
6653 enabled/used. @var{flags} initially may have either the SCHED_RGN or SCHED_EBB
6654 bit set. This denotes the scheduler pass for which the data should be
6655 provided. The target backend should modify @var{flags} by modifying
6656 the bits corresponding to the following features: USE_DEPS_LIST, USE_GLAT,
6657 DETACH_LIFE_INFO, and DO_SPECULATION@. For the DO_SPECULATION feature
6658 an additional structure @var{spec_info} should be filled by the target.
6659 The structure describes speculation types that can be used in the scheduler.
6662 @deftypefn {Target Hook} int TARGET_SCHED_SMS_RES_MII (struct ddg *@var{g})
6663 This hook is called by the swing modulo scheduler to calculate a
6664 resource-based lower bound which is based on the resources available in
6665 the machine and the resources required by each instruction. The target
6666 backend can use @var{g} to calculate such bound. A very simple lower
6667 bound will be used in case this hook is not implemented: the total number
6668 of instructions divided by the issue rate.
6672 @section Dividing the Output into Sections (Texts, Data, @dots{})
6673 @c the above section title is WAY too long. maybe cut the part between
6674 @c the (...)? --mew 10feb93
6676 An object file is divided into sections containing different types of
6677 data. In the most common case, there are three sections: the @dfn{text
6678 section}, which holds instructions and read-only data; the @dfn{data
6679 section}, which holds initialized writable data; and the @dfn{bss
6680 section}, which holds uninitialized data. Some systems have other kinds
6683 @file{varasm.c} provides several well-known sections, such as
6684 @code{text_section}, @code{data_section} and @code{bss_section}.
6685 The normal way of controlling a @code{@var{foo}_section} variable
6686 is to define the associated @code{@var{FOO}_SECTION_ASM_OP} macro,
6687 as described below. The macros are only read once, when @file{varasm.c}
6688 initializes itself, so their values must be run-time constants.
6689 They may however depend on command-line flags.
6691 @emph{Note:} Some run-time files, such @file{crtstuff.c}, also make
6692 use of the @code{@var{FOO}_SECTION_ASM_OP} macros, and expect them
6693 to be string literals.
6695 Some assemblers require a different string to be written every time a
6696 section is selected. If your assembler falls into this category, you
6697 should define the @code{TARGET_ASM_INIT_SECTIONS} hook and use
6698 @code{get_unnamed_section} to set up the sections.
6700 You must always create a @code{text_section}, either by defining
6701 @code{TEXT_SECTION_ASM_OP} or by initializing @code{text_section}
6702 in @code{TARGET_ASM_INIT_SECTIONS}. The same is true of
6703 @code{data_section} and @code{DATA_SECTION_ASM_OP}. If you do not
6704 create a distinct @code{readonly_data_section}, the default is to
6705 reuse @code{text_section}.
6707 All the other @file{varasm.c} sections are optional, and are null
6708 if the target does not provide them.
6710 @defmac TEXT_SECTION_ASM_OP
6711 A C expression whose value is a string, including spacing, containing the
6712 assembler operation that should precede instructions and read-only data.
6713 Normally @code{"\t.text"} is right.
6716 @defmac HOT_TEXT_SECTION_NAME
6717 If defined, a C string constant for the name of the section containing most
6718 frequently executed functions of the program. If not defined, GCC will provide
6719 a default definition if the target supports named sections.
6722 @defmac UNLIKELY_EXECUTED_TEXT_SECTION_NAME
6723 If defined, a C string constant for the name of the section containing unlikely
6724 executed functions in the program.
6727 @defmac DATA_SECTION_ASM_OP
6728 A C expression whose value is a string, including spacing, containing the
6729 assembler operation to identify the following data as writable initialized
6730 data. Normally @code{"\t.data"} is right.
6733 @defmac SDATA_SECTION_ASM_OP
6734 If defined, a C expression whose value is a string, including spacing,
6735 containing the assembler operation to identify the following data as
6736 initialized, writable small data.
6739 @defmac READONLY_DATA_SECTION_ASM_OP
6740 A C expression whose value is a string, including spacing, containing the
6741 assembler operation to identify the following data as read-only initialized
6745 @defmac BSS_SECTION_ASM_OP
6746 If defined, a C expression whose value is a string, including spacing,
6747 containing the assembler operation to identify the following data as
6748 uninitialized global data. If not defined, and neither
6749 @code{ASM_OUTPUT_BSS} nor @code{ASM_OUTPUT_ALIGNED_BSS} are defined,
6750 uninitialized global data will be output in the data section if
6751 @option{-fno-common} is passed, otherwise @code{ASM_OUTPUT_COMMON} will be
6755 @defmac SBSS_SECTION_ASM_OP
6756 If defined, a C expression whose value is a string, including spacing,
6757 containing the assembler operation to identify the following data as
6758 uninitialized, writable small data.
6761 @defmac INIT_SECTION_ASM_OP
6762 If defined, a C expression whose value is a string, including spacing,
6763 containing the assembler operation to identify the following data as
6764 initialization code. If not defined, GCC will assume such a section does
6765 not exist. This section has no corresponding @code{init_section}
6766 variable; it is used entirely in runtime code.
6769 @defmac FINI_SECTION_ASM_OP
6770 If defined, a C expression whose value is a string, including spacing,
6771 containing the assembler operation to identify the following data as
6772 finalization code. If not defined, GCC will assume such a section does
6773 not exist. This section has no corresponding @code{fini_section}
6774 variable; it is used entirely in runtime code.
6777 @defmac INIT_ARRAY_SECTION_ASM_OP
6778 If defined, a C expression whose value is a string, including spacing,
6779 containing the assembler operation to identify the following data as
6780 part of the @code{.init_array} (or equivalent) section. If not
6781 defined, GCC will assume such a section does not exist. Do not define
6782 both this macro and @code{INIT_SECTION_ASM_OP}.
6785 @defmac FINI_ARRAY_SECTION_ASM_OP
6786 If defined, a C expression whose value is a string, including spacing,
6787 containing the assembler operation to identify the following data as
6788 part of the @code{.fini_array} (or equivalent) section. If not
6789 defined, GCC will assume such a section does not exist. Do not define
6790 both this macro and @code{FINI_SECTION_ASM_OP}.
6793 @defmac CRT_CALL_STATIC_FUNCTION (@var{section_op}, @var{function})
6794 If defined, an ASM statement that switches to a different section
6795 via @var{section_op}, calls @var{function}, and switches back to
6796 the text section. This is used in @file{crtstuff.c} if
6797 @code{INIT_SECTION_ASM_OP} or @code{FINI_SECTION_ASM_OP} to calls
6798 to initialization and finalization functions from the init and fini
6799 sections. By default, this macro uses a simple function call. Some
6800 ports need hand-crafted assembly code to avoid dependencies on
6801 registers initialized in the function prologue or to ensure that
6802 constant pools don't end up too far way in the text section.
6805 @defmac TARGET_LIBGCC_SDATA_SECTION
6806 If defined, a string which names the section into which small
6807 variables defined in crtstuff and libgcc should go. This is useful
6808 when the target has options for optimizing access to small data, and
6809 you want the crtstuff and libgcc routines to be conservative in what
6810 they expect of your application yet liberal in what your application
6811 expects. For example, for targets with a @code{.sdata} section (like
6812 MIPS), you could compile crtstuff with @code{-G 0} so that it doesn't
6813 require small data support from your application, but use this macro
6814 to put small data into @code{.sdata} so that your application can
6815 access these variables whether it uses small data or not.
6818 @defmac FORCE_CODE_SECTION_ALIGN
6819 If defined, an ASM statement that aligns a code section to some
6820 arbitrary boundary. This is used to force all fragments of the
6821 @code{.init} and @code{.fini} sections to have to same alignment
6822 and thus prevent the linker from having to add any padding.
6825 @defmac JUMP_TABLES_IN_TEXT_SECTION
6826 Define this macro to be an expression with a nonzero value if jump
6827 tables (for @code{tablejump} insns) should be output in the text
6828 section, along with the assembler instructions. Otherwise, the
6829 readonly data section is used.
6831 This macro is irrelevant if there is no separate readonly data section.
6834 @deftypefn {Target Hook} void TARGET_ASM_INIT_SECTIONS (void)
6835 Define this hook if you need to do something special to set up the
6836 @file{varasm.c} sections, or if your target has some special sections
6837 of its own that you need to create.
6839 GCC calls this hook after processing the command line, but before writing
6840 any assembly code, and before calling any of the section-returning hooks
6844 @deftypefn {Target Hook} TARGET_ASM_RELOC_RW_MASK (void)
6845 Return a mask describing how relocations should be treated when
6846 selecting sections. Bit 1 should be set if global relocations
6847 should be placed in a read-write section; bit 0 should be set if
6848 local relocations should be placed in a read-write section.
6850 The default version of this function returns 3 when @option{-fpic}
6851 is in effect, and 0 otherwise. The hook is typically redefined
6852 when the target cannot support (some kinds of) dynamic relocations
6853 in read-only sections even in executables.
6856 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_SECTION (tree @var{exp}, int @var{reloc}, unsigned HOST_WIDE_INT @var{align})
6857 Return the section into which @var{exp} should be placed. You can
6858 assume that @var{exp} is either a @code{VAR_DECL} node or a constant of
6859 some sort. @var{reloc} indicates whether the initial value of @var{exp}
6860 requires link-time relocations. Bit 0 is set when variable contains
6861 local relocations only, while bit 1 is set for global relocations.
6862 @var{align} is the constant alignment in bits.
6864 The default version of this function takes care of putting read-only
6865 variables in @code{readonly_data_section}.
6867 See also @var{USE_SELECT_SECTION_FOR_FUNCTIONS}.
6870 @defmac USE_SELECT_SECTION_FOR_FUNCTIONS
6871 Define this macro if you wish TARGET_ASM_SELECT_SECTION to be called
6872 for @code{FUNCTION_DECL}s as well as for variables and constants.
6874 In the case of a @code{FUNCTION_DECL}, @var{reloc} will be zero if the
6875 function has been determined to be likely to be called, and nonzero if
6876 it is unlikely to be called.
6879 @deftypefn {Target Hook} void TARGET_ASM_UNIQUE_SECTION (tree @var{decl}, int @var{reloc})
6880 Build up a unique section name, expressed as a @code{STRING_CST} node,
6881 and assign it to @samp{DECL_SECTION_NAME (@var{decl})}.
6882 As with @code{TARGET_ASM_SELECT_SECTION}, @var{reloc} indicates whether
6883 the initial value of @var{exp} requires link-time relocations.
6885 The default version of this function appends the symbol name to the
6886 ELF section name that would normally be used for the symbol. For
6887 example, the function @code{foo} would be placed in @code{.text.foo}.
6888 Whatever the actual target object format, this is often good enough.
6891 @deftypefn {Target Hook} {section *} TARGET_ASM_FUNCTION_RODATA_SECTION (tree @var{decl})
6892 Return the readonly data section associated with
6893 @samp{DECL_SECTION_NAME (@var{decl})}.
6894 The default version of this function selects @code{.gnu.linkonce.r.name} if
6895 the function's section is @code{.gnu.linkonce.t.name}, @code{.rodata.name}
6896 if function is in @code{.text.name}, and the normal readonly-data section
6900 @deftypefn {Target Hook} {section *} TARGET_ASM_SELECT_RTX_SECTION (enum machine_mode @var{mode}, rtx @var{x}, unsigned HOST_WIDE_INT @var{align})
6901 Return the section into which a constant @var{x}, of mode @var{mode},
6902 should be placed. You can assume that @var{x} is some kind of
6903 constant in RTL@. The argument @var{mode} is redundant except in the
6904 case of a @code{const_int} rtx. @var{align} is the constant alignment
6907 The default version of this function takes care of putting symbolic
6908 constants in @code{flag_pic} mode in @code{data_section} and everything
6909 else in @code{readonly_data_section}.
6912 @deftypefn {Target Hook} void TARGET_MANGLE_DECL_ASSEMBLER_NAME (tree @var{decl}, tree @var{id})
6913 Define this hook if you need to postprocess the assembler name generated
6914 by target-independent code. The @var{id} provided to this hook will be
6915 the computed name (e.g., the macro @code{DECL_NAME} of the @var{decl} in C,
6916 or the mangled name of the @var{decl} in C++). The return value of the
6917 hook is an @code{IDENTIFIER_NODE} for the appropriate mangled name on
6918 your target system. The default implementation of this hook just
6919 returns the @var{id} provided.
6922 @deftypefn {Target Hook} void TARGET_ENCODE_SECTION_INFO (tree @var{decl}, rtx @var{rtl}, int @var{new_decl_p})
6923 Define this hook if references to a symbol or a constant must be
6924 treated differently depending on something about the variable or
6925 function named by the symbol (such as what section it is in).
6927 The hook is executed immediately after rtl has been created for
6928 @var{decl}, which may be a variable or function declaration or
6929 an entry in the constant pool. In either case, @var{rtl} is the
6930 rtl in question. Do @emph{not} use @code{DECL_RTL (@var{decl})}
6931 in this hook; that field may not have been initialized yet.
6933 In the case of a constant, it is safe to assume that the rtl is
6934 a @code{mem} whose address is a @code{symbol_ref}. Most decls
6935 will also have this form, but that is not guaranteed. Global
6936 register variables, for instance, will have a @code{reg} for their
6937 rtl. (Normally the right thing to do with such unusual rtl is
6940 The @var{new_decl_p} argument will be true if this is the first time
6941 that @code{TARGET_ENCODE_SECTION_INFO} has been invoked on this decl. It will
6942 be false for subsequent invocations, which will happen for duplicate
6943 declarations. Whether or not anything must be done for the duplicate
6944 declaration depends on whether the hook examines @code{DECL_ATTRIBUTES}.
6945 @var{new_decl_p} is always true when the hook is called for a constant.
6947 @cindex @code{SYMBOL_REF_FLAG}, in @code{TARGET_ENCODE_SECTION_INFO}
6948 The usual thing for this hook to do is to record flags in the
6949 @code{symbol_ref}, using @code{SYMBOL_REF_FLAG} or @code{SYMBOL_REF_FLAGS}.
6950 Historically, the name string was modified if it was necessary to
6951 encode more than one bit of information, but this practice is now
6952 discouraged; use @code{SYMBOL_REF_FLAGS}.
6954 The default definition of this hook, @code{default_encode_section_info}
6955 in @file{varasm.c}, sets a number of commonly-useful bits in
6956 @code{SYMBOL_REF_FLAGS}. Check whether the default does what you need
6957 before overriding it.
6960 @deftypefn {Target Hook} const char *TARGET_STRIP_NAME_ENCODING (const char *name)
6961 Decode @var{name} and return the real name part, sans
6962 the characters that @code{TARGET_ENCODE_SECTION_INFO}
6966 @deftypefn {Target Hook} bool TARGET_IN_SMALL_DATA_P (tree @var{exp})
6967 Returns true if @var{exp} should be placed into a ``small data'' section.
6968 The default version of this hook always returns false.
6971 @deftypevr {Target Hook} bool TARGET_HAVE_SRODATA_SECTION
6972 Contains the value true if the target places read-only
6973 ``small data'' into a separate section. The default value is false.
6976 @deftypefn {Target Hook} bool TARGET_BINDS_LOCAL_P (tree @var{exp})
6977 Returns true if @var{exp} names an object for which name resolution
6978 rules must resolve to the current ``module'' (dynamic shared library
6979 or executable image).
6981 The default version of this hook implements the name resolution rules
6982 for ELF, which has a looser model of global name binding than other
6983 currently supported object file formats.
6986 @deftypevr {Target Hook} bool TARGET_HAVE_TLS
6987 Contains the value true if the target supports thread-local storage.
6988 The default value is false.
6993 @section Position Independent Code
6994 @cindex position independent code
6997 This section describes macros that help implement generation of position
6998 independent code. Simply defining these macros is not enough to
6999 generate valid PIC; you must also add support to the hook
7000 @code{TARGET_LEGITIMATE_ADDRESS_P} and to the macro
7001 @code{PRINT_OPERAND_ADDRESS}, as well as @code{LEGITIMIZE_ADDRESS}. You
7002 must modify the definition of @samp{movsi} to do something appropriate
7003 when the source operand contains a symbolic address. You may also
7004 need to alter the handling of switch statements so that they use
7006 @c i rearranged the order of the macros above to try to force one of
7007 @c them to the next line, to eliminate an overfull hbox. --mew 10feb93
7009 @defmac PIC_OFFSET_TABLE_REGNUM
7010 The register number of the register used to address a table of static
7011 data addresses in memory. In some cases this register is defined by a
7012 processor's ``application binary interface'' (ABI)@. When this macro
7013 is defined, RTL is generated for this register once, as with the stack
7014 pointer and frame pointer registers. If this macro is not defined, it
7015 is up to the machine-dependent files to allocate such a register (if
7016 necessary). Note that this register must be fixed when in use (e.g.@:
7017 when @code{flag_pic} is true).
7020 @defmac PIC_OFFSET_TABLE_REG_CALL_CLOBBERED
7021 Define this macro if the register defined by
7022 @code{PIC_OFFSET_TABLE_REGNUM} is clobbered by calls. Do not define
7023 this macro if @code{PIC_OFFSET_TABLE_REGNUM} is not defined.
7026 @defmac LEGITIMATE_PIC_OPERAND_P (@var{x})
7027 A C expression that is nonzero if @var{x} is a legitimate immediate
7028 operand on the target machine when generating position independent code.
7029 You can assume that @var{x} satisfies @code{CONSTANT_P}, so you need not
7030 check this. You can also assume @var{flag_pic} is true, so you need not
7031 check it either. You need not define this macro if all constants
7032 (including @code{SYMBOL_REF}) can be immediate operands when generating
7033 position independent code.
7036 @node Assembler Format
7037 @section Defining the Output Assembler Language
7039 This section describes macros whose principal purpose is to describe how
7040 to write instructions in assembler language---rather than what the
7044 * File Framework:: Structural information for the assembler file.
7045 * Data Output:: Output of constants (numbers, strings, addresses).
7046 * Uninitialized Data:: Output of uninitialized variables.
7047 * Label Output:: Output and generation of labels.
7048 * Initialization:: General principles of initialization
7049 and termination routines.
7050 * Macros for Initialization::
7051 Specific macros that control the handling of
7052 initialization and termination routines.
7053 * Instruction Output:: Output of actual instructions.
7054 * Dispatch Tables:: Output of jump tables.
7055 * Exception Region Output:: Output of exception region code.
7056 * Alignment Output:: Pseudo ops for alignment and skipping data.
7059 @node File Framework
7060 @subsection The Overall Framework of an Assembler File
7061 @cindex assembler format
7062 @cindex output of assembler code
7064 @c prevent bad page break with this line
7065 This describes the overall framework of an assembly file.
7067 @deftypefn {Target Hook} void TARGET_ASM_FILE_START ()
7068 @findex default_file_start
7069 Output to @code{asm_out_file} any text which the assembler expects to
7070 find at the beginning of a file. The default behavior is controlled
7071 by two flags, documented below. Unless your target's assembler is
7072 quite unusual, if you override the default, you should call
7073 @code{default_file_start} at some point in your target hook. This
7074 lets other target files rely on these variables.
7077 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_APP_OFF
7078 If this flag is true, the text of the macro @code{ASM_APP_OFF} will be
7079 printed as the very first line in the assembly file, unless
7080 @option{-fverbose-asm} is in effect. (If that macro has been defined
7081 to the empty string, this variable has no effect.) With the normal
7082 definition of @code{ASM_APP_OFF}, the effect is to notify the GNU
7083 assembler that it need not bother stripping comments or extra
7084 whitespace from its input. This allows it to work a bit faster.
7086 The default is false. You should not set it to true unless you have
7087 verified that your port does not generate any extra whitespace or
7088 comments that will cause GAS to issue errors in NO_APP mode.
7091 @deftypevr {Target Hook} bool TARGET_ASM_FILE_START_FILE_DIRECTIVE
7092 If this flag is true, @code{output_file_directive} will be called
7093 for the primary source file, immediately after printing
7094 @code{ASM_APP_OFF} (if that is enabled). Most ELF assemblers expect
7095 this to be done. The default is false.
7098 @deftypefn {Target Hook} void TARGET_ASM_FILE_END ()
7099 Output to @code{asm_out_file} any text which the assembler expects
7100 to find at the end of a file. The default is to output nothing.
7103 @deftypefun void file_end_indicate_exec_stack ()
7104 Some systems use a common convention, the @samp{.note.GNU-stack}
7105 special section, to indicate whether or not an object file relies on
7106 the stack being executable. If your system uses this convention, you
7107 should define @code{TARGET_ASM_FILE_END} to this function. If you
7108 need to do other things in that hook, have your hook function call
7112 @defmac ASM_COMMENT_START
7113 A C string constant describing how to begin a comment in the target
7114 assembler language. The compiler assumes that the comment will end at
7115 the end of the line.
7119 A C string constant for text to be output before each @code{asm}
7120 statement or group of consecutive ones. Normally this is
7121 @code{"#APP"}, which is a comment that has no effect on most
7122 assemblers but tells the GNU assembler that it must check the lines
7123 that follow for all valid assembler constructs.
7127 A C string constant for text to be output after each @code{asm}
7128 statement or group of consecutive ones. Normally this is
7129 @code{"#NO_APP"}, which tells the GNU assembler to resume making the
7130 time-saving assumptions that are valid for ordinary compiler output.
7133 @defmac ASM_OUTPUT_SOURCE_FILENAME (@var{stream}, @var{name})
7134 A C statement to output COFF information or DWARF debugging information
7135 which indicates that filename @var{name} is the current source file to
7136 the stdio stream @var{stream}.
7138 This macro need not be defined if the standard form of output
7139 for the file format in use is appropriate.
7142 @defmac OUTPUT_QUOTED_STRING (@var{stream}, @var{string})
7143 A C statement to output the string @var{string} to the stdio stream
7144 @var{stream}. If you do not call the function @code{output_quoted_string}
7145 in your config files, GCC will only call it to output filenames to
7146 the assembler source. So you can use it to canonicalize the format
7147 of the filename using this macro.
7150 @defmac ASM_OUTPUT_IDENT (@var{stream}, @var{string})
7151 A C statement to output something to the assembler file to handle a
7152 @samp{#ident} directive containing the text @var{string}. If this
7153 macro is not defined, nothing is output for a @samp{#ident} directive.
7156 @deftypefn {Target Hook} void TARGET_ASM_NAMED_SECTION (const char *@var{name}, unsigned int @var{flags}, unsigned int @var{align})
7157 Output assembly directives to switch to section @var{name}. The section
7158 should have attributes as specified by @var{flags}, which is a bit mask
7159 of the @code{SECTION_*} flags defined in @file{output.h}. If @var{align}
7160 is nonzero, it contains an alignment in bytes to be used for the section,
7161 otherwise some target default should be used. Only targets that must
7162 specify an alignment within the section directive need pay attention to
7163 @var{align} -- we will still use @code{ASM_OUTPUT_ALIGN}.
7166 @deftypefn {Target Hook} bool TARGET_HAVE_NAMED_SECTIONS
7167 This flag is true if the target supports @code{TARGET_ASM_NAMED_SECTION}.
7170 @anchor{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}
7171 @deftypefn {Target Hook} bool TARGET_HAVE_SWITCHABLE_BSS_SECTIONS
7172 This flag is true if we can create zeroed data by switching to a BSS
7173 section and then using @code{ASM_OUTPUT_SKIP} to allocate the space.
7174 This is true on most ELF targets.
7177 @deftypefn {Target Hook} {unsigned int} TARGET_SECTION_TYPE_FLAGS (tree @var{decl}, const char *@var{name}, int @var{reloc})
7178 Choose a set of section attributes for use by @code{TARGET_ASM_NAMED_SECTION}
7179 based on a variable or function decl, a section name, and whether or not the
7180 declaration's initializer may contain runtime relocations. @var{decl} may be
7181 null, in which case read-write data should be assumed.
7183 The default version of this function handles choosing code vs data,
7184 read-only vs read-write data, and @code{flag_pic}. You should only
7185 need to override this if your target has special flags that might be
7186 set via @code{__attribute__}.
7189 @deftypefn {Target Hook} {int} TARGET_ASM_RECORD_GCC_SWITCHES (print_switch_type @var{type}, const char * @var{text})
7190 Provides the target with the ability to record the gcc command line
7191 switches that have been passed to the compiler, and options that are
7192 enabled. The @var{type} argument specifies what is being recorded.
7193 It can take the following values:
7196 @item SWITCH_TYPE_PASSED
7197 @var{text} is a command line switch that has been set by the user.
7199 @item SWITCH_TYPE_ENABLED
7200 @var{text} is an option which has been enabled. This might be as a
7201 direct result of a command line switch, or because it is enabled by
7202 default or because it has been enabled as a side effect of a different
7203 command line switch. For example, the @option{-O2} switch enables
7204 various different individual optimization passes.
7206 @item SWITCH_TYPE_DESCRIPTIVE
7207 @var{text} is either NULL or some descriptive text which should be
7208 ignored. If @var{text} is NULL then it is being used to warn the
7209 target hook that either recording is starting or ending. The first
7210 time @var{type} is SWITCH_TYPE_DESCRIPTIVE and @var{text} is NULL, the
7211 warning is for start up and the second time the warning is for
7212 wind down. This feature is to allow the target hook to make any
7213 necessary preparations before it starts to record switches and to
7214 perform any necessary tidying up after it has finished recording
7217 @item SWITCH_TYPE_LINE_START
7218 This option can be ignored by this target hook.
7220 @item SWITCH_TYPE_LINE_END
7221 This option can be ignored by this target hook.
7224 The hook's return value must be zero. Other return values may be
7225 supported in the future.
7227 By default this hook is set to NULL, but an example implementation is
7228 provided for ELF based targets. Called @var{elf_record_gcc_switches},
7229 it records the switches as ASCII text inside a new, string mergeable
7230 section in the assembler output file. The name of the new section is
7231 provided by the @code{TARGET_ASM_RECORD_GCC_SWITCHES_SECTION} target
7235 @deftypefn {Target Hook} {const char *} TARGET_ASM_RECORD_GCC_SWITCHES_SECTION
7236 This is the name of the section that will be created by the example
7237 ELF implementation of the @code{TARGET_ASM_RECORD_GCC_SWITCHES} target
7243 @subsection Output of Data
7246 @deftypevr {Target Hook} {const char *} TARGET_ASM_BYTE_OP
7247 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_HI_OP
7248 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_SI_OP
7249 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_DI_OP
7250 @deftypevrx {Target Hook} {const char *} TARGET_ASM_ALIGNED_TI_OP
7251 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_HI_OP
7252 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_SI_OP
7253 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_DI_OP
7254 @deftypevrx {Target Hook} {const char *} TARGET_ASM_UNALIGNED_TI_OP
7255 These hooks specify assembly directives for creating certain kinds
7256 of integer object. The @code{TARGET_ASM_BYTE_OP} directive creates a
7257 byte-sized object, the @code{TARGET_ASM_ALIGNED_HI_OP} one creates an
7258 aligned two-byte object, and so on. Any of the hooks may be
7259 @code{NULL}, indicating that no suitable directive is available.
7261 The compiler will print these strings at the start of a new line,
7262 followed immediately by the object's initial value. In most cases,
7263 the string should contain a tab, a pseudo-op, and then another tab.
7266 @deftypefn {Target Hook} bool TARGET_ASM_INTEGER (rtx @var{x}, unsigned int @var{size}, int @var{aligned_p})
7267 The @code{assemble_integer} function uses this hook to output an
7268 integer object. @var{x} is the object's value, @var{size} is its size
7269 in bytes and @var{aligned_p} indicates whether it is aligned. The
7270 function should return @code{true} if it was able to output the
7271 object. If it returns false, @code{assemble_integer} will try to
7272 split the object into smaller parts.
7274 The default implementation of this hook will use the
7275 @code{TARGET_ASM_BYTE_OP} family of strings, returning @code{false}
7276 when the relevant string is @code{NULL}.
7279 @defmac OUTPUT_ADDR_CONST_EXTRA (@var{stream}, @var{x}, @var{fail})
7280 A C statement to recognize @var{rtx} patterns that
7281 @code{output_addr_const} can't deal with, and output assembly code to
7282 @var{stream} corresponding to the pattern @var{x}. This may be used to
7283 allow machine-dependent @code{UNSPEC}s to appear within constants.
7285 If @code{OUTPUT_ADDR_CONST_EXTRA} fails to recognize a pattern, it must
7286 @code{goto fail}, so that a standard error message is printed. If it
7287 prints an error message itself, by calling, for example,
7288 @code{output_operand_lossage}, it may just complete normally.
7291 @defmac ASM_OUTPUT_ASCII (@var{stream}, @var{ptr}, @var{len})
7292 A C statement to output to the stdio stream @var{stream} an assembler
7293 instruction to assemble a string constant containing the @var{len}
7294 bytes at @var{ptr}. @var{ptr} will be a C expression of type
7295 @code{char *} and @var{len} a C expression of type @code{int}.
7297 If the assembler has a @code{.ascii} pseudo-op as found in the
7298 Berkeley Unix assembler, do not define the macro
7299 @code{ASM_OUTPUT_ASCII}.
7302 @defmac ASM_OUTPUT_FDESC (@var{stream}, @var{decl}, @var{n})
7303 A C statement to output word @var{n} of a function descriptor for
7304 @var{decl}. This must be defined if @code{TARGET_VTABLE_USES_DESCRIPTORS}
7305 is defined, and is otherwise unused.
7308 @defmac CONSTANT_POOL_BEFORE_FUNCTION
7309 You may define this macro as a C expression. You should define the
7310 expression to have a nonzero value if GCC should output the constant
7311 pool for a function before the code for the function, or a zero value if
7312 GCC should output the constant pool after the function. If you do
7313 not define this macro, the usual case, GCC will output the constant
7314 pool before the function.
7317 @defmac ASM_OUTPUT_POOL_PROLOGUE (@var{file}, @var{funname}, @var{fundecl}, @var{size})
7318 A C statement to output assembler commands to define the start of the
7319 constant pool for a function. @var{funname} is a string giving
7320 the name of the function. Should the return type of the function
7321 be required, it can be obtained via @var{fundecl}. @var{size}
7322 is the size, in bytes, of the constant pool that will be written
7323 immediately after this call.
7325 If no constant-pool prefix is required, the usual case, this macro need
7329 @defmac ASM_OUTPUT_SPECIAL_POOL_ENTRY (@var{file}, @var{x}, @var{mode}, @var{align}, @var{labelno}, @var{jumpto})
7330 A C statement (with or without semicolon) to output a constant in the
7331 constant pool, if it needs special treatment. (This macro need not do
7332 anything for RTL expressions that can be output normally.)
7334 The argument @var{file} is the standard I/O stream to output the
7335 assembler code on. @var{x} is the RTL expression for the constant to
7336 output, and @var{mode} is the machine mode (in case @var{x} is a
7337 @samp{const_int}). @var{align} is the required alignment for the value
7338 @var{x}; you should output an assembler directive to force this much
7341 The argument @var{labelno} is a number to use in an internal label for
7342 the address of this pool entry. The definition of this macro is
7343 responsible for outputting the label definition at the proper place.
7344 Here is how to do this:
7347 @code{(*targetm.asm_out.internal_label)} (@var{file}, "LC", @var{labelno});
7350 When you output a pool entry specially, you should end with a
7351 @code{goto} to the label @var{jumpto}. This will prevent the same pool
7352 entry from being output a second time in the usual manner.
7354 You need not define this macro if it would do nothing.
7357 @defmac ASM_OUTPUT_POOL_EPILOGUE (@var{file} @var{funname} @var{fundecl} @var{size})
7358 A C statement to output assembler commands to at the end of the constant
7359 pool for a function. @var{funname} is a string giving the name of the
7360 function. Should the return type of the function be required, you can
7361 obtain it via @var{fundecl}. @var{size} is the size, in bytes, of the
7362 constant pool that GCC wrote immediately before this call.
7364 If no constant-pool epilogue is required, the usual case, you need not
7368 @defmac IS_ASM_LOGICAL_LINE_SEPARATOR (@var{C}, @var{STR})
7369 Define this macro as a C expression which is nonzero if @var{C} is
7370 used as a logical line separator by the assembler. @var{STR} points
7371 to the position in the string where @var{C} was found; this can be used if
7372 a line separator uses multiple characters.
7374 If you do not define this macro, the default is that only
7375 the character @samp{;} is treated as a logical line separator.
7378 @deftypevr {Target Hook} {const char *} TARGET_ASM_OPEN_PAREN
7379 @deftypevrx {Target Hook} {const char *} TARGET_ASM_CLOSE_PAREN
7380 These target hooks are C string constants, describing the syntax in the
7381 assembler for grouping arithmetic expressions. If not overridden, they
7382 default to normal parentheses, which is correct for most assemblers.
7385 These macros are provided by @file{real.h} for writing the definitions
7386 of @code{ASM_OUTPUT_DOUBLE} and the like:
7388 @defmac REAL_VALUE_TO_TARGET_SINGLE (@var{x}, @var{l})
7389 @defmacx REAL_VALUE_TO_TARGET_DOUBLE (@var{x}, @var{l})
7390 @defmacx REAL_VALUE_TO_TARGET_LONG_DOUBLE (@var{x}, @var{l})
7391 @defmacx REAL_VALUE_TO_TARGET_DECIMAL32 (@var{x}, @var{l})
7392 @defmacx REAL_VALUE_TO_TARGET_DECIMAL64 (@var{x}, @var{l})
7393 @defmacx REAL_VALUE_TO_TARGET_DECIMAL128 (@var{x}, @var{l})
7394 These translate @var{x}, of type @code{REAL_VALUE_TYPE}, to the
7395 target's floating point representation, and store its bit pattern in
7396 the variable @var{l}. For @code{REAL_VALUE_TO_TARGET_SINGLE} and
7397 @code{REAL_VALUE_TO_TARGET_DECIMAL32}, this variable should be a
7398 simple @code{long int}. For the others, it should be an array of
7399 @code{long int}. The number of elements in this array is determined
7400 by the size of the desired target floating point data type: 32 bits of
7401 it go in each @code{long int} array element. Each array element holds
7402 32 bits of the result, even if @code{long int} is wider than 32 bits
7403 on the host machine.
7405 The array element values are designed so that you can print them out
7406 using @code{fprintf} in the order they should appear in the target
7410 @node Uninitialized Data
7411 @subsection Output of Uninitialized Variables
7413 Each of the macros in this section is used to do the whole job of
7414 outputting a single uninitialized variable.
7416 @defmac ASM_OUTPUT_COMMON (@var{stream}, @var{name}, @var{size}, @var{rounded})
7417 A C statement (sans semicolon) to output to the stdio stream
7418 @var{stream} the assembler definition of a common-label named
7419 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7420 is the size rounded up to whatever alignment the caller wants. It is
7421 possible that @var{size} may be zero, for instance if a struct with no
7422 other member than a zero-length array is defined. In this case, the
7423 backend must output a symbol definition that allocates at least one
7424 byte, both so that the address of the resulting object does not compare
7425 equal to any other, and because some object formats cannot even express
7426 the concept of a zero-sized common symbol, as that is how they represent
7427 an ordinary undefined external.
7429 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7430 output the name itself; before and after that, output the additional
7431 assembler syntax for defining the name, and a newline.
7433 This macro controls how the assembler definitions of uninitialized
7434 common global variables are output.
7437 @defmac ASM_OUTPUT_ALIGNED_COMMON (@var{stream}, @var{name}, @var{size}, @var{alignment})
7438 Like @code{ASM_OUTPUT_COMMON} except takes the required alignment as a
7439 separate, explicit argument. If you define this macro, it is used in
7440 place of @code{ASM_OUTPUT_COMMON}, and gives you more flexibility in
7441 handling the required alignment of the variable. The alignment is specified
7442 as the number of bits.
7445 @defmac ASM_OUTPUT_ALIGNED_DECL_COMMON (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7446 Like @code{ASM_OUTPUT_ALIGNED_COMMON} except that @var{decl} of the
7447 variable to be output, if there is one, or @code{NULL_TREE} if there
7448 is no corresponding variable. If you define this macro, GCC will use it
7449 in place of both @code{ASM_OUTPUT_COMMON} and
7450 @code{ASM_OUTPUT_ALIGNED_COMMON}. Define this macro when you need to see
7451 the variable's decl in order to chose what to output.
7454 @defmac ASM_OUTPUT_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{rounded})
7455 A C statement (sans semicolon) to output to the stdio stream
7456 @var{stream} the assembler definition of uninitialized global @var{decl} named
7457 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7458 is the size rounded up to whatever alignment the caller wants.
7460 Try to use function @code{asm_output_bss} defined in @file{varasm.c} when
7461 defining this macro. If unable, use the expression
7462 @code{assemble_name (@var{stream}, @var{name})} to output the name itself;
7463 before and after that, output the additional assembler syntax for defining
7464 the name, and a newline.
7466 There are two ways of handling global BSS@. One is to define either
7467 this macro or its aligned counterpart, @code{ASM_OUTPUT_ALIGNED_BSS}.
7468 The other is to have @code{TARGET_ASM_SELECT_SECTION} return a
7469 switchable BSS section (@pxref{TARGET_HAVE_SWITCHABLE_BSS_SECTIONS}).
7470 You do not need to do both.
7472 Some languages do not have @code{common} data, and require a
7473 non-common form of global BSS in order to handle uninitialized globals
7474 efficiently. C++ is one example of this. However, if the target does
7475 not support global BSS, the front end may choose to make globals
7476 common in order to save space in the object file.
7479 @defmac ASM_OUTPUT_ALIGNED_BSS (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7480 Like @code{ASM_OUTPUT_BSS} except takes the required alignment as a
7481 separate, explicit argument. If you define this macro, it is used in
7482 place of @code{ASM_OUTPUT_BSS}, and gives you more flexibility in
7483 handling the required alignment of the variable. The alignment is specified
7484 as the number of bits.
7486 Try to use function @code{asm_output_aligned_bss} defined in file
7487 @file{varasm.c} when defining this macro.
7490 @defmac ASM_OUTPUT_LOCAL (@var{stream}, @var{name}, @var{size}, @var{rounded})
7491 A C statement (sans semicolon) to output to the stdio stream
7492 @var{stream} the assembler definition of a local-common-label named
7493 @var{name} whose size is @var{size} bytes. The variable @var{rounded}
7494 is the size rounded up to whatever alignment the caller wants.
7496 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7497 output the name itself; before and after that, output the additional
7498 assembler syntax for defining the name, and a newline.
7500 This macro controls how the assembler definitions of uninitialized
7501 static variables are output.
7504 @defmac ASM_OUTPUT_ALIGNED_LOCAL (@var{stream}, @var{name}, @var{size}, @var{alignment})
7505 Like @code{ASM_OUTPUT_LOCAL} except takes the required alignment as a
7506 separate, explicit argument. If you define this macro, it is used in
7507 place of @code{ASM_OUTPUT_LOCAL}, and gives you more flexibility in
7508 handling the required alignment of the variable. The alignment is specified
7509 as the number of bits.
7512 @defmac ASM_OUTPUT_ALIGNED_DECL_LOCAL (@var{stream}, @var{decl}, @var{name}, @var{size}, @var{alignment})
7513 Like @code{ASM_OUTPUT_ALIGNED_DECL} except that @var{decl} of the
7514 variable to be output, if there is one, or @code{NULL_TREE} if there
7515 is no corresponding variable. If you define this macro, GCC will use it
7516 in place of both @code{ASM_OUTPUT_DECL} and
7517 @code{ASM_OUTPUT_ALIGNED_DECL}. Define this macro when you need to see
7518 the variable's decl in order to chose what to output.
7522 @subsection Output and Generation of Labels
7524 @c prevent bad page break with this line
7525 This is about outputting labels.
7527 @findex assemble_name
7528 @defmac ASM_OUTPUT_LABEL (@var{stream}, @var{name})
7529 A C statement (sans semicolon) to output to the stdio stream
7530 @var{stream} the assembler definition of a label named @var{name}.
7531 Use the expression @code{assemble_name (@var{stream}, @var{name})} to
7532 output the name itself; before and after that, output the additional
7533 assembler syntax for defining the name, and a newline. A default
7534 definition of this macro is provided which is correct for most systems.
7537 @findex assemble_name_raw
7538 @defmac ASM_OUTPUT_INTERNAL_LABEL (@var{stream}, @var{name})
7539 Identical to @code{ASM_OUTPUT_LABEL}, except that @var{name} is known
7540 to refer to a compiler-generated label. The default definition uses
7541 @code{assemble_name_raw}, which is like @code{assemble_name} except
7542 that it is more efficient.
7546 A C string containing the appropriate assembler directive to specify the
7547 size of a symbol, without any arguments. On systems that use ELF, the
7548 default (in @file{config/elfos.h}) is @samp{"\t.size\t"}; on other
7549 systems, the default is not to define this macro.
7551 Define this macro only if it is correct to use the default definitions
7552 of @code{ASM_OUTPUT_SIZE_DIRECTIVE} and @code{ASM_OUTPUT_MEASURED_SIZE}
7553 for your system. If you need your own custom definitions of those
7554 macros, or if you do not need explicit symbol sizes at all, do not
7558 @defmac ASM_OUTPUT_SIZE_DIRECTIVE (@var{stream}, @var{name}, @var{size})
7559 A C statement (sans semicolon) to output to the stdio stream
7560 @var{stream} a directive telling the assembler that the size of the
7561 symbol @var{name} is @var{size}. @var{size} is a @code{HOST_WIDE_INT}.
7562 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7566 @defmac ASM_OUTPUT_MEASURED_SIZE (@var{stream}, @var{name})
7567 A C statement (sans semicolon) to output to the stdio stream
7568 @var{stream} a directive telling the assembler to calculate the size of
7569 the symbol @var{name} by subtracting its address from the current
7572 If you define @code{SIZE_ASM_OP}, a default definition of this macro is
7573 provided. The default assumes that the assembler recognizes a special
7574 @samp{.} symbol as referring to the current address, and can calculate
7575 the difference between this and another symbol. If your assembler does
7576 not recognize @samp{.} or cannot do calculations with it, you will need
7577 to redefine @code{ASM_OUTPUT_MEASURED_SIZE} to use some other technique.
7581 A C string containing the appropriate assembler directive to specify the
7582 type of a symbol, without any arguments. On systems that use ELF, the
7583 default (in @file{config/elfos.h}) is @samp{"\t.type\t"}; on other
7584 systems, the default is not to define this macro.
7586 Define this macro only if it is correct to use the default definition of
7587 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7588 custom definition of this macro, or if you do not need explicit symbol
7589 types at all, do not define this macro.
7592 @defmac TYPE_OPERAND_FMT
7593 A C string which specifies (using @code{printf} syntax) the format of
7594 the second operand to @code{TYPE_ASM_OP}. On systems that use ELF, the
7595 default (in @file{config/elfos.h}) is @samp{"@@%s"}; on other systems,
7596 the default is not to define this macro.
7598 Define this macro only if it is correct to use the default definition of
7599 @code{ASM_OUTPUT_TYPE_DIRECTIVE} for your system. If you need your own
7600 custom definition of this macro, or if you do not need explicit symbol
7601 types at all, do not define this macro.
7604 @defmac ASM_OUTPUT_TYPE_DIRECTIVE (@var{stream}, @var{type})
7605 A C statement (sans semicolon) to output to the stdio stream
7606 @var{stream} a directive telling the assembler that the type of the
7607 symbol @var{name} is @var{type}. @var{type} is a C string; currently,
7608 that string is always either @samp{"function"} or @samp{"object"}, but
7609 you should not count on this.
7611 If you define @code{TYPE_ASM_OP} and @code{TYPE_OPERAND_FMT}, a default
7612 definition of this macro is provided.
7615 @defmac ASM_DECLARE_FUNCTION_NAME (@var{stream}, @var{name}, @var{decl})
7616 A C statement (sans semicolon) to output to the stdio stream
7617 @var{stream} any text necessary for declaring the name @var{name} of a
7618 function which is being defined. This macro is responsible for
7619 outputting the label definition (perhaps using
7620 @code{ASM_OUTPUT_LABEL}). The argument @var{decl} is the
7621 @code{FUNCTION_DECL} tree node representing the function.
7623 If this macro is not defined, then the function name is defined in the
7624 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7626 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7630 @defmac ASM_DECLARE_FUNCTION_SIZE (@var{stream}, @var{name}, @var{decl})
7631 A C statement (sans semicolon) to output to the stdio stream
7632 @var{stream} any text necessary for declaring the size of a function
7633 which is being defined. The argument @var{name} is the name of the
7634 function. The argument @var{decl} is the @code{FUNCTION_DECL} tree node
7635 representing the function.
7637 If this macro is not defined, then the function size is not defined.
7639 You may wish to use @code{ASM_OUTPUT_MEASURED_SIZE} in the definition
7643 @defmac ASM_DECLARE_OBJECT_NAME (@var{stream}, @var{name}, @var{decl})
7644 A C statement (sans semicolon) to output to the stdio stream
7645 @var{stream} any text necessary for declaring the name @var{name} of an
7646 initialized variable which is being defined. This macro must output the
7647 label definition (perhaps using @code{ASM_OUTPUT_LABEL}). The argument
7648 @var{decl} is the @code{VAR_DECL} tree node representing the variable.
7650 If this macro is not defined, then the variable name is defined in the
7651 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7653 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} and/or
7654 @code{ASM_OUTPUT_SIZE_DIRECTIVE} in the definition of this macro.
7657 @defmac ASM_DECLARE_CONSTANT_NAME (@var{stream}, @var{name}, @var{exp}, @var{size})
7658 A C statement (sans semicolon) to output to the stdio stream
7659 @var{stream} any text necessary for declaring the name @var{name} of a
7660 constant which is being defined. This macro is responsible for
7661 outputting the label definition (perhaps using
7662 @code{ASM_OUTPUT_LABEL}). The argument @var{exp} is the
7663 value of the constant, and @var{size} is the size of the constant
7664 in bytes. @var{name} will be an internal label.
7666 If this macro is not defined, then the @var{name} is defined in the
7667 usual manner as a label (by means of @code{ASM_OUTPUT_LABEL}).
7669 You may wish to use @code{ASM_OUTPUT_TYPE_DIRECTIVE} in the definition
7673 @defmac ASM_DECLARE_REGISTER_GLOBAL (@var{stream}, @var{decl}, @var{regno}, @var{name})
7674 A C statement (sans semicolon) to output to the stdio stream
7675 @var{stream} any text necessary for claiming a register @var{regno}
7676 for a global variable @var{decl} with name @var{name}.
7678 If you don't define this macro, that is equivalent to defining it to do
7682 @defmac ASM_FINISH_DECLARE_OBJECT (@var{stream}, @var{decl}, @var{toplevel}, @var{atend})
7683 A C statement (sans semicolon) to finish up declaring a variable name
7684 once the compiler has processed its initializer fully and thus has had a
7685 chance to determine the size of an array when controlled by an
7686 initializer. This is used on systems where it's necessary to declare
7687 something about the size of the object.
7689 If you don't define this macro, that is equivalent to defining it to do
7692 You may wish to use @code{ASM_OUTPUT_SIZE_DIRECTIVE} and/or
7693 @code{ASM_OUTPUT_MEASURED_SIZE} in the definition of this macro.
7696 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_LABEL (FILE *@var{stream}, const char *@var{name})
7697 This target hook is a function to output to the stdio stream
7698 @var{stream} some commands that will make the label @var{name} global;
7699 that is, available for reference from other files.
7701 The default implementation relies on a proper definition of
7702 @code{GLOBAL_ASM_OP}.
7705 @deftypefn {Target Hook} void TARGET_ASM_GLOBALIZE_DECL_NAME (FILE *@var{stream}, tree @var{decl})
7706 This target hook is a function to output to the stdio stream
7707 @var{stream} some commands that will make the name associated with @var{decl}
7708 global; that is, available for reference from other files.
7710 The default implementation uses the TARGET_ASM_GLOBALIZE_LABEL target hook.
7713 @defmac ASM_WEAKEN_LABEL (@var{stream}, @var{name})
7714 A C statement (sans semicolon) to output to the stdio stream
7715 @var{stream} some commands that will make the label @var{name} weak;
7716 that is, available for reference from other files but only used if
7717 no other definition is available. Use the expression
7718 @code{assemble_name (@var{stream}, @var{name})} to output the name
7719 itself; before and after that, output the additional assembler syntax
7720 for making that name weak, and a newline.
7722 If you don't define this macro or @code{ASM_WEAKEN_DECL}, GCC will not
7723 support weak symbols and you should not define the @code{SUPPORTS_WEAK}
7727 @defmac ASM_WEAKEN_DECL (@var{stream}, @var{decl}, @var{name}, @var{value})
7728 Combines (and replaces) the function of @code{ASM_WEAKEN_LABEL} and
7729 @code{ASM_OUTPUT_WEAK_ALIAS}, allowing access to the associated function
7730 or variable decl. If @var{value} is not @code{NULL}, this C statement
7731 should output to the stdio stream @var{stream} assembler code which
7732 defines (equates) the weak symbol @var{name} to have the value
7733 @var{value}. If @var{value} is @code{NULL}, it should output commands
7734 to make @var{name} weak.
7737 @defmac ASM_OUTPUT_WEAKREF (@var{stream}, @var{decl}, @var{name}, @var{value})
7738 Outputs a directive that enables @var{name} to be used to refer to
7739 symbol @var{value} with weak-symbol semantics. @code{decl} is the
7740 declaration of @code{name}.
7743 @defmac SUPPORTS_WEAK
7744 A C expression which evaluates to true if the target supports weak symbols.
7746 If you don't define this macro, @file{defaults.h} provides a default
7747 definition. If either @code{ASM_WEAKEN_LABEL} or @code{ASM_WEAKEN_DECL}
7748 is defined, the default definition is @samp{1}; otherwise, it is
7749 @samp{0}. Define this macro if you want to control weak symbol support
7750 with a compiler flag such as @option{-melf}.
7753 @defmac MAKE_DECL_ONE_ONLY (@var{decl})
7754 A C statement (sans semicolon) to mark @var{decl} to be emitted as a
7755 public symbol such that extra copies in multiple translation units will
7756 be discarded by the linker. Define this macro if your object file
7757 format provides support for this concept, such as the @samp{COMDAT}
7758 section flags in the Microsoft Windows PE/COFF format, and this support
7759 requires changes to @var{decl}, such as putting it in a separate section.
7762 @defmac SUPPORTS_ONE_ONLY
7763 A C expression which evaluates to true if the target supports one-only
7766 If you don't define this macro, @file{varasm.c} provides a default
7767 definition. If @code{MAKE_DECL_ONE_ONLY} is defined, the default
7768 definition is @samp{1}; otherwise, it is @samp{0}. Define this macro if
7769 you want to control one-only symbol support with a compiler flag, or if
7770 setting the @code{DECL_ONE_ONLY} flag is enough to mark a declaration to
7771 be emitted as one-only.
7774 @deftypefn {Target Hook} void TARGET_ASM_ASSEMBLE_VISIBILITY (tree @var{decl}, const char *@var{visibility})
7775 This target hook is a function to output to @var{asm_out_file} some
7776 commands that will make the symbol(s) associated with @var{decl} have
7777 hidden, protected or internal visibility as specified by @var{visibility}.
7780 @defmac TARGET_WEAK_NOT_IN_ARCHIVE_TOC
7781 A C expression that evaluates to true if the target's linker expects
7782 that weak symbols do not appear in a static archive's table of contents.
7783 The default is @code{0}.
7785 Leaving weak symbols out of an archive's table of contents means that,
7786 if a symbol will only have a definition in one translation unit and
7787 will have undefined references from other translation units, that
7788 symbol should not be weak. Defining this macro to be nonzero will
7789 thus have the effect that certain symbols that would normally be weak
7790 (explicit template instantiations, and vtables for polymorphic classes
7791 with noninline key methods) will instead be nonweak.
7793 The C++ ABI requires this macro to be zero. Define this macro for
7794 targets where full C++ ABI compliance is impossible and where linker
7795 restrictions require weak symbols to be left out of a static archive's
7799 @defmac ASM_OUTPUT_EXTERNAL (@var{stream}, @var{decl}, @var{name})
7800 A C statement (sans semicolon) to output to the stdio stream
7801 @var{stream} any text necessary for declaring the name of an external
7802 symbol named @var{name} which is referenced in this compilation but
7803 not defined. The value of @var{decl} is the tree node for the
7806 This macro need not be defined if it does not need to output anything.
7807 The GNU assembler and most Unix assemblers don't require anything.
7810 @deftypefn {Target Hook} void TARGET_ASM_EXTERNAL_LIBCALL (rtx @var{symref})
7811 This target hook is a function to output to @var{asm_out_file} an assembler
7812 pseudo-op to declare a library function name external. The name of the
7813 library function is given by @var{symref}, which is a @code{symbol_ref}.
7816 @deftypefn {Target Hook} void TARGET_ASM_MARK_DECL_PRESERVED (tree @var{decl})
7817 This target hook is a function to output to @var{asm_out_file} an assembler
7818 directive to annotate used symbol. Darwin target use .no_dead_code_strip
7822 @defmac ASM_OUTPUT_LABELREF (@var{stream}, @var{name})
7823 A C statement (sans semicolon) to output to the stdio stream
7824 @var{stream} a reference in assembler syntax to a label named
7825 @var{name}. This should add @samp{_} to the front of the name, if that
7826 is customary on your operating system, as it is in most Berkeley Unix
7827 systems. This macro is used in @code{assemble_name}.
7830 @defmac ASM_OUTPUT_SYMBOL_REF (@var{stream}, @var{sym})
7831 A C statement (sans semicolon) to output a reference to
7832 @code{SYMBOL_REF} @var{sym}. If not defined, @code{assemble_name}
7833 will be used to output the name of the symbol. This macro may be used
7834 to modify the way a symbol is referenced depending on information
7835 encoded by @code{TARGET_ENCODE_SECTION_INFO}.
7838 @defmac ASM_OUTPUT_LABEL_REF (@var{stream}, @var{buf})
7839 A C statement (sans semicolon) to output a reference to @var{buf}, the
7840 result of @code{ASM_GENERATE_INTERNAL_LABEL}. If not defined,
7841 @code{assemble_name} will be used to output the name of the symbol.
7842 This macro is not used by @code{output_asm_label}, or the @code{%l}
7843 specifier that calls it; the intention is that this macro should be set
7844 when it is necessary to output a label differently when its address is
7848 @deftypefn {Target Hook} void TARGET_ASM_INTERNAL_LABEL (FILE *@var{stream}, const char *@var{prefix}, unsigned long @var{labelno})
7849 A function to output to the stdio stream @var{stream} a label whose
7850 name is made from the string @var{prefix} and the number @var{labelno}.
7852 It is absolutely essential that these labels be distinct from the labels
7853 used for user-level functions and variables. Otherwise, certain programs
7854 will have name conflicts with internal labels.
7856 It is desirable to exclude internal labels from the symbol table of the
7857 object file. Most assemblers have a naming convention for labels that
7858 should be excluded; on many systems, the letter @samp{L} at the
7859 beginning of a label has this effect. You should find out what
7860 convention your system uses, and follow it.
7862 The default version of this function utilizes @code{ASM_GENERATE_INTERNAL_LABEL}.
7865 @defmac ASM_OUTPUT_DEBUG_LABEL (@var{stream}, @var{prefix}, @var{num})
7866 A C statement to output to the stdio stream @var{stream} a debug info
7867 label whose name is made from the string @var{prefix} and the number
7868 @var{num}. This is useful for VLIW targets, where debug info labels
7869 may need to be treated differently than branch target labels. On some
7870 systems, branch target labels must be at the beginning of instruction
7871 bundles, but debug info labels can occur in the middle of instruction
7874 If this macro is not defined, then @code{(*targetm.asm_out.internal_label)} will be
7878 @defmac ASM_GENERATE_INTERNAL_LABEL (@var{string}, @var{prefix}, @var{num})
7879 A C statement to store into the string @var{string} a label whose name
7880 is made from the string @var{prefix} and the number @var{num}.
7882 This string, when output subsequently by @code{assemble_name}, should
7883 produce the output that @code{(*targetm.asm_out.internal_label)} would produce
7884 with the same @var{prefix} and @var{num}.
7886 If the string begins with @samp{*}, then @code{assemble_name} will
7887 output the rest of the string unchanged. It is often convenient for
7888 @code{ASM_GENERATE_INTERNAL_LABEL} to use @samp{*} in this way. If the
7889 string doesn't start with @samp{*}, then @code{ASM_OUTPUT_LABELREF} gets
7890 to output the string, and may change it. (Of course,
7891 @code{ASM_OUTPUT_LABELREF} is also part of your machine description, so
7892 you should know what it does on your machine.)
7895 @defmac ASM_FORMAT_PRIVATE_NAME (@var{outvar}, @var{name}, @var{number})
7896 A C expression to assign to @var{outvar} (which is a variable of type
7897 @code{char *}) a newly allocated string made from the string
7898 @var{name} and the number @var{number}, with some suitable punctuation
7899 added. Use @code{alloca} to get space for the string.
7901 The string will be used as an argument to @code{ASM_OUTPUT_LABELREF} to
7902 produce an assembler label for an internal static variable whose name is
7903 @var{name}. Therefore, the string must be such as to result in valid
7904 assembler code. The argument @var{number} is different each time this
7905 macro is executed; it prevents conflicts between similarly-named
7906 internal static variables in different scopes.
7908 Ideally this string should not be a valid C identifier, to prevent any
7909 conflict with the user's own symbols. Most assemblers allow periods
7910 or percent signs in assembler symbols; putting at least one of these
7911 between the name and the number will suffice.
7913 If this macro is not defined, a default definition will be provided
7914 which is correct for most systems.
7917 @defmac ASM_OUTPUT_DEF (@var{stream}, @var{name}, @var{value})
7918 A C statement to output to the stdio stream @var{stream} assembler code
7919 which defines (equates) the symbol @var{name} to have the value @var{value}.
7922 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7923 correct for most systems.
7926 @defmac ASM_OUTPUT_DEF_FROM_DECLS (@var{stream}, @var{decl_of_name}, @var{decl_of_value})
7927 A C statement to output to the stdio stream @var{stream} assembler code
7928 which defines (equates) the symbol whose tree node is @var{decl_of_name}
7929 to have the value of the tree node @var{decl_of_value}. This macro will
7930 be used in preference to @samp{ASM_OUTPUT_DEF} if it is defined and if
7931 the tree nodes are available.
7934 If @code{SET_ASM_OP} is defined, a default definition is provided which is
7935 correct for most systems.
7938 @defmac TARGET_DEFERRED_OUTPUT_DEFS (@var{decl_of_name}, @var{decl_of_value})
7939 A C statement that evaluates to true if the assembler code which defines
7940 (equates) the symbol whose tree node is @var{decl_of_name} to have the value
7941 of the tree node @var{decl_of_value} should be emitted near the end of the
7942 current compilation unit. The default is to not defer output of defines.
7943 This macro affects defines output by @samp{ASM_OUTPUT_DEF} and
7944 @samp{ASM_OUTPUT_DEF_FROM_DECLS}.
7947 @defmac ASM_OUTPUT_WEAK_ALIAS (@var{stream}, @var{name}, @var{value})
7948 A C statement to output to the stdio stream @var{stream} assembler code
7949 which defines (equates) the weak symbol @var{name} to have the value
7950 @var{value}. If @var{value} is @code{NULL}, it defines @var{name} as
7951 an undefined weak symbol.
7953 Define this macro if the target only supports weak aliases; define
7954 @code{ASM_OUTPUT_DEF} instead if possible.
7957 @defmac OBJC_GEN_METHOD_LABEL (@var{buf}, @var{is_inst}, @var{class_name}, @var{cat_name}, @var{sel_name})
7958 Define this macro to override the default assembler names used for
7959 Objective-C methods.
7961 The default name is a unique method number followed by the name of the
7962 class (e.g.@: @samp{_1_Foo}). For methods in categories, the name of
7963 the category is also included in the assembler name (e.g.@:
7966 These names are safe on most systems, but make debugging difficult since
7967 the method's selector is not present in the name. Therefore, particular
7968 systems define other ways of computing names.
7970 @var{buf} is an expression of type @code{char *} which gives you a
7971 buffer in which to store the name; its length is as long as
7972 @var{class_name}, @var{cat_name} and @var{sel_name} put together, plus
7973 50 characters extra.
7975 The argument @var{is_inst} specifies whether the method is an instance
7976 method or a class method; @var{class_name} is the name of the class;
7977 @var{cat_name} is the name of the category (or @code{NULL} if the method is not
7978 in a category); and @var{sel_name} is the name of the selector.
7980 On systems where the assembler can handle quoted names, you can use this
7981 macro to provide more human-readable names.
7984 @defmac ASM_DECLARE_CLASS_REFERENCE (@var{stream}, @var{name})
7985 A C statement (sans semicolon) to output to the stdio stream
7986 @var{stream} commands to declare that the label @var{name} is an
7987 Objective-C class reference. This is only needed for targets whose
7988 linkers have special support for NeXT-style runtimes.
7991 @defmac ASM_DECLARE_UNRESOLVED_REFERENCE (@var{stream}, @var{name})
7992 A C statement (sans semicolon) to output to the stdio stream
7993 @var{stream} commands to declare that the label @var{name} is an
7994 unresolved Objective-C class reference. This is only needed for targets
7995 whose linkers have special support for NeXT-style runtimes.
7998 @node Initialization
7999 @subsection How Initialization Functions Are Handled
8000 @cindex initialization routines
8001 @cindex termination routines
8002 @cindex constructors, output of
8003 @cindex destructors, output of
8005 The compiled code for certain languages includes @dfn{constructors}
8006 (also called @dfn{initialization routines})---functions to initialize
8007 data in the program when the program is started. These functions need
8008 to be called before the program is ``started''---that is to say, before
8009 @code{main} is called.
8011 Compiling some languages generates @dfn{destructors} (also called
8012 @dfn{termination routines}) that should be called when the program
8015 To make the initialization and termination functions work, the compiler
8016 must output something in the assembler code to cause those functions to
8017 be called at the appropriate time. When you port the compiler to a new
8018 system, you need to specify how to do this.
8020 There are two major ways that GCC currently supports the execution of
8021 initialization and termination functions. Each way has two variants.
8022 Much of the structure is common to all four variations.
8024 @findex __CTOR_LIST__
8025 @findex __DTOR_LIST__
8026 The linker must build two lists of these functions---a list of
8027 initialization functions, called @code{__CTOR_LIST__}, and a list of
8028 termination functions, called @code{__DTOR_LIST__}.
8030 Each list always begins with an ignored function pointer (which may hold
8031 0, @minus{}1, or a count of the function pointers after it, depending on
8032 the environment). This is followed by a series of zero or more function
8033 pointers to constructors (or destructors), followed by a function
8034 pointer containing zero.
8036 Depending on the operating system and its executable file format, either
8037 @file{crtstuff.c} or @file{libgcc2.c} traverses these lists at startup
8038 time and exit time. Constructors are called in reverse order of the
8039 list; destructors in forward order.
8041 The best way to handle static constructors works only for object file
8042 formats which provide arbitrarily-named sections. A section is set
8043 aside for a list of constructors, and another for a list of destructors.
8044 Traditionally these are called @samp{.ctors} and @samp{.dtors}. Each
8045 object file that defines an initialization function also puts a word in
8046 the constructor section to point to that function. The linker
8047 accumulates all these words into one contiguous @samp{.ctors} section.
8048 Termination functions are handled similarly.
8050 This method will be chosen as the default by @file{target-def.h} if
8051 @code{TARGET_ASM_NAMED_SECTION} is defined. A target that does not
8052 support arbitrary sections, but does support special designated
8053 constructor and destructor sections may define @code{CTORS_SECTION_ASM_OP}
8054 and @code{DTORS_SECTION_ASM_OP} to achieve the same effect.
8056 When arbitrary sections are available, there are two variants, depending
8057 upon how the code in @file{crtstuff.c} is called. On systems that
8058 support a @dfn{.init} section which is executed at program startup,
8059 parts of @file{crtstuff.c} are compiled into that section. The
8060 program is linked by the @command{gcc} driver like this:
8063 ld -o @var{output_file} crti.o crtbegin.o @dots{} -lgcc crtend.o crtn.o
8066 The prologue of a function (@code{__init}) appears in the @code{.init}
8067 section of @file{crti.o}; the epilogue appears in @file{crtn.o}. Likewise
8068 for the function @code{__fini} in the @dfn{.fini} section. Normally these
8069 files are provided by the operating system or by the GNU C library, but
8070 are provided by GCC for a few targets.
8072 The objects @file{crtbegin.o} and @file{crtend.o} are (for most targets)
8073 compiled from @file{crtstuff.c}. They contain, among other things, code
8074 fragments within the @code{.init} and @code{.fini} sections that branch
8075 to routines in the @code{.text} section. The linker will pull all parts
8076 of a section together, which results in a complete @code{__init} function
8077 that invokes the routines we need at startup.
8079 To use this variant, you must define the @code{INIT_SECTION_ASM_OP}
8082 If no init section is available, when GCC compiles any function called
8083 @code{main} (or more accurately, any function designated as a program
8084 entry point by the language front end calling @code{expand_main_function}),
8085 it inserts a procedure call to @code{__main} as the first executable code
8086 after the function prologue. The @code{__main} function is defined
8087 in @file{libgcc2.c} and runs the global constructors.
8089 In file formats that don't support arbitrary sections, there are again
8090 two variants. In the simplest variant, the GNU linker (GNU @code{ld})
8091 and an `a.out' format must be used. In this case,
8092 @code{TARGET_ASM_CONSTRUCTOR} is defined to produce a @code{.stabs}
8093 entry of type @samp{N_SETT}, referencing the name @code{__CTOR_LIST__},
8094 and with the address of the void function containing the initialization
8095 code as its value. The GNU linker recognizes this as a request to add
8096 the value to a @dfn{set}; the values are accumulated, and are eventually
8097 placed in the executable as a vector in the format described above, with
8098 a leading (ignored) count and a trailing zero element.
8099 @code{TARGET_ASM_DESTRUCTOR} is handled similarly. Since no init
8100 section is available, the absence of @code{INIT_SECTION_ASM_OP} causes
8101 the compilation of @code{main} to call @code{__main} as above, starting
8102 the initialization process.
8104 The last variant uses neither arbitrary sections nor the GNU linker.
8105 This is preferable when you want to do dynamic linking and when using
8106 file formats which the GNU linker does not support, such as `ECOFF'@. In
8107 this case, @code{TARGET_HAVE_CTORS_DTORS} is false, initialization and
8108 termination functions are recognized simply by their names. This requires
8109 an extra program in the linkage step, called @command{collect2}. This program
8110 pretends to be the linker, for use with GCC; it does its job by running
8111 the ordinary linker, but also arranges to include the vectors of
8112 initialization and termination functions. These functions are called
8113 via @code{__main} as described above. In order to use this method,
8114 @code{use_collect2} must be defined in the target in @file{config.gcc}.
8117 The following section describes the specific macros that control and
8118 customize the handling of initialization and termination functions.
8121 @node Macros for Initialization
8122 @subsection Macros Controlling Initialization Routines
8124 Here are the macros that control how the compiler handles initialization
8125 and termination functions:
8127 @defmac INIT_SECTION_ASM_OP
8128 If defined, a C string constant, including spacing, for the assembler
8129 operation to identify the following data as initialization code. If not
8130 defined, GCC will assume such a section does not exist. When you are
8131 using special sections for initialization and termination functions, this
8132 macro also controls how @file{crtstuff.c} and @file{libgcc2.c} arrange to
8133 run the initialization functions.
8136 @defmac HAS_INIT_SECTION
8137 If defined, @code{main} will not call @code{__main} as described above.
8138 This macro should be defined for systems that control start-up code
8139 on a symbol-by-symbol basis, such as OSF/1, and should not
8140 be defined explicitly for systems that support @code{INIT_SECTION_ASM_OP}.
8143 @defmac LD_INIT_SWITCH
8144 If defined, a C string constant for a switch that tells the linker that
8145 the following symbol is an initialization routine.
8148 @defmac LD_FINI_SWITCH
8149 If defined, a C string constant for a switch that tells the linker that
8150 the following symbol is a finalization routine.
8153 @defmac COLLECT_SHARED_INIT_FUNC (@var{stream}, @var{func})
8154 If defined, a C statement that will write a function that can be
8155 automatically called when a shared library is loaded. The function
8156 should call @var{func}, which takes no arguments. If not defined, and
8157 the object format requires an explicit initialization function, then a
8158 function called @code{_GLOBAL__DI} will be generated.
8160 This function and the following one are used by collect2 when linking a
8161 shared library that needs constructors or destructors, or has DWARF2
8162 exception tables embedded in the code.
8165 @defmac COLLECT_SHARED_FINI_FUNC (@var{stream}, @var{func})
8166 If defined, a C statement that will write a function that can be
8167 automatically called when a shared library is unloaded. The function
8168 should call @var{func}, which takes no arguments. If not defined, and
8169 the object format requires an explicit finalization function, then a
8170 function called @code{_GLOBAL__DD} will be generated.
8173 @defmac INVOKE__main
8174 If defined, @code{main} will call @code{__main} despite the presence of
8175 @code{INIT_SECTION_ASM_OP}. This macro should be defined for systems
8176 where the init section is not actually run automatically, but is still
8177 useful for collecting the lists of constructors and destructors.
8180 @defmac SUPPORTS_INIT_PRIORITY
8181 If nonzero, the C++ @code{init_priority} attribute is supported and the
8182 compiler should emit instructions to control the order of initialization
8183 of objects. If zero, the compiler will issue an error message upon
8184 encountering an @code{init_priority} attribute.
8187 @deftypefn {Target Hook} bool TARGET_HAVE_CTORS_DTORS
8188 This value is true if the target supports some ``native'' method of
8189 collecting constructors and destructors to be run at startup and exit.
8190 It is false if we must use @command{collect2}.
8193 @deftypefn {Target Hook} void TARGET_ASM_CONSTRUCTOR (rtx @var{symbol}, int @var{priority})
8194 If defined, a function that outputs assembler code to arrange to call
8195 the function referenced by @var{symbol} at initialization time.
8197 Assume that @var{symbol} is a @code{SYMBOL_REF} for a function taking
8198 no arguments and with no return value. If the target supports initialization
8199 priorities, @var{priority} is a value between 0 and @code{MAX_INIT_PRIORITY};
8200 otherwise it must be @code{DEFAULT_INIT_PRIORITY}.
8202 If this macro is not defined by the target, a suitable default will
8203 be chosen if (1) the target supports arbitrary section names, (2) the
8204 target defines @code{CTORS_SECTION_ASM_OP}, or (3) @code{USE_COLLECT2}
8208 @deftypefn {Target Hook} void TARGET_ASM_DESTRUCTOR (rtx @var{symbol}, int @var{priority})
8209 This is like @code{TARGET_ASM_CONSTRUCTOR} but used for termination
8210 functions rather than initialization functions.
8213 If @code{TARGET_HAVE_CTORS_DTORS} is true, the initialization routine
8214 generated for the generated object file will have static linkage.
8216 If your system uses @command{collect2} as the means of processing
8217 constructors, then that program normally uses @command{nm} to scan
8218 an object file for constructor functions to be called.
8220 On certain kinds of systems, you can define this macro to make
8221 @command{collect2} work faster (and, in some cases, make it work at all):
8223 @defmac OBJECT_FORMAT_COFF
8224 Define this macro if the system uses COFF (Common Object File Format)
8225 object files, so that @command{collect2} can assume this format and scan
8226 object files directly for dynamic constructor/destructor functions.
8228 This macro is effective only in a native compiler; @command{collect2} as
8229 part of a cross compiler always uses @command{nm} for the target machine.
8232 @defmac REAL_NM_FILE_NAME
8233 Define this macro as a C string constant containing the file name to use
8234 to execute @command{nm}. The default is to search the path normally for
8237 If your system supports shared libraries and has a program to list the
8238 dynamic dependencies of a given library or executable, you can define
8239 these macros to enable support for running initialization and
8240 termination functions in shared libraries:
8244 Define this macro to a C string constant containing the name of the program
8245 which lists dynamic dependencies, like @command{"ldd"} under SunOS 4.
8248 @defmac PARSE_LDD_OUTPUT (@var{ptr})
8249 Define this macro to be C code that extracts filenames from the output
8250 of the program denoted by @code{LDD_SUFFIX}. @var{ptr} is a variable
8251 of type @code{char *} that points to the beginning of a line of output
8252 from @code{LDD_SUFFIX}. If the line lists a dynamic dependency, the
8253 code must advance @var{ptr} to the beginning of the filename on that
8254 line. Otherwise, it must set @var{ptr} to @code{NULL}.
8257 @defmac SHLIB_SUFFIX
8258 Define this macro to a C string constant containing the default shared
8259 library extension of the target (e.g., @samp{".so"}). @command{collect2}
8260 strips version information after this suffix when generating global
8261 constructor and destructor names. This define is only needed on targets
8262 that use @command{collect2} to process constructors and destructors.
8265 @node Instruction Output
8266 @subsection Output of Assembler Instructions
8268 @c prevent bad page break with this line
8269 This describes assembler instruction output.
8271 @defmac REGISTER_NAMES
8272 A C initializer containing the assembler's names for the machine
8273 registers, each one as a C string constant. This is what translates
8274 register numbers in the compiler into assembler language.
8277 @defmac ADDITIONAL_REGISTER_NAMES
8278 If defined, a C initializer for an array of structures containing a name
8279 and a register number. This macro defines additional names for hard
8280 registers, thus allowing the @code{asm} option in declarations to refer
8281 to registers using alternate names.
8284 @defmac ASM_OUTPUT_OPCODE (@var{stream}, @var{ptr})
8285 Define this macro if you are using an unusual assembler that
8286 requires different names for the machine instructions.
8288 The definition is a C statement or statements which output an
8289 assembler instruction opcode to the stdio stream @var{stream}. The
8290 macro-operand @var{ptr} is a variable of type @code{char *} which
8291 points to the opcode name in its ``internal'' form---the form that is
8292 written in the machine description. The definition should output the
8293 opcode name to @var{stream}, performing any translation you desire, and
8294 increment the variable @var{ptr} to point at the end of the opcode
8295 so that it will not be output twice.
8297 In fact, your macro definition may process less than the entire opcode
8298 name, or more than the opcode name; but if you want to process text
8299 that includes @samp{%}-sequences to substitute operands, you must take
8300 care of the substitution yourself. Just be sure to increment
8301 @var{ptr} over whatever text should not be output normally.
8303 @findex recog_data.operand
8304 If you need to look at the operand values, they can be found as the
8305 elements of @code{recog_data.operand}.
8307 If the macro definition does nothing, the instruction is output
8311 @defmac FINAL_PRESCAN_INSN (@var{insn}, @var{opvec}, @var{noperands})
8312 If defined, a C statement to be executed just prior to the output of
8313 assembler code for @var{insn}, to modify the extracted operands so
8314 they will be output differently.
8316 Here the argument @var{opvec} is the vector containing the operands
8317 extracted from @var{insn}, and @var{noperands} is the number of
8318 elements of the vector which contain meaningful data for this insn.
8319 The contents of this vector are what will be used to convert the insn
8320 template into assembler code, so you can change the assembler output
8321 by changing the contents of the vector.
8323 This macro is useful when various assembler syntaxes share a single
8324 file of instruction patterns; by defining this macro differently, you
8325 can cause a large class of instructions to be output differently (such
8326 as with rearranged operands). Naturally, variations in assembler
8327 syntax affecting individual insn patterns ought to be handled by
8328 writing conditional output routines in those patterns.
8330 If this macro is not defined, it is equivalent to a null statement.
8333 @deftypefn {Target Hook} void TARGET_ASM_FINAL_POSTSCAN_INSN (FILE *@var{FILE}, rtx @var{insn}, rtx *@var{opvec}, int @var{noperands})
8334 If defined, this target hook is a function which is executed just after the
8335 output of assembler code for @var{insn}, to change the mode of the assembler
8338 Here the argument @var{opvec} is the vector containing the operands
8339 extracted from @var{insn}, and @var{noperands} is the number of
8340 elements of the vector which contain meaningful data for this insn.
8341 The contents of this vector are what was used to convert the insn
8342 template into assembler code, so you can change the assembler mode
8343 by checking the contents of the vector.
8346 @defmac PRINT_OPERAND (@var{stream}, @var{x}, @var{code})
8347 A C compound statement to output to stdio stream @var{stream} the
8348 assembler syntax for an instruction operand @var{x}. @var{x} is an
8351 @var{code} is a value that can be used to specify one of several ways
8352 of printing the operand. It is used when identical operands must be
8353 printed differently depending on the context. @var{code} comes from
8354 the @samp{%} specification that was used to request printing of the
8355 operand. If the specification was just @samp{%@var{digit}} then
8356 @var{code} is 0; if the specification was @samp{%@var{ltr}
8357 @var{digit}} then @var{code} is the ASCII code for @var{ltr}.
8360 If @var{x} is a register, this macro should print the register's name.
8361 The names can be found in an array @code{reg_names} whose type is
8362 @code{char *[]}. @code{reg_names} is initialized from
8363 @code{REGISTER_NAMES}.
8365 When the machine description has a specification @samp{%@var{punct}}
8366 (a @samp{%} followed by a punctuation character), this macro is called
8367 with a null pointer for @var{x} and the punctuation character for
8371 @defmac PRINT_OPERAND_PUNCT_VALID_P (@var{code})
8372 A C expression which evaluates to true if @var{code} is a valid
8373 punctuation character for use in the @code{PRINT_OPERAND} macro. If
8374 @code{PRINT_OPERAND_PUNCT_VALID_P} is not defined, it means that no
8375 punctuation characters (except for the standard one, @samp{%}) are used
8379 @defmac PRINT_OPERAND_ADDRESS (@var{stream}, @var{x})
8380 A C compound statement to output to stdio stream @var{stream} the
8381 assembler syntax for an instruction operand that is a memory reference
8382 whose address is @var{x}. @var{x} is an RTL expression.
8384 @cindex @code{TARGET_ENCODE_SECTION_INFO} usage
8385 On some machines, the syntax for a symbolic address depends on the
8386 section that the address refers to. On these machines, define the hook
8387 @code{TARGET_ENCODE_SECTION_INFO} to store the information into the
8388 @code{symbol_ref}, and then check for it here. @xref{Assembler
8392 @findex dbr_sequence_length
8393 @defmac DBR_OUTPUT_SEQEND (@var{file})
8394 A C statement, to be executed after all slot-filler instructions have
8395 been output. If necessary, call @code{dbr_sequence_length} to
8396 determine the number of slots filled in a sequence (zero if not
8397 currently outputting a sequence), to decide how many no-ops to output,
8400 Don't define this macro if it has nothing to do, but it is helpful in
8401 reading assembly output if the extent of the delay sequence is made
8402 explicit (e.g.@: with white space).
8405 @findex final_sequence
8406 Note that output routines for instructions with delay slots must be
8407 prepared to deal with not being output as part of a sequence
8408 (i.e.@: when the scheduling pass is not run, or when no slot fillers could be
8409 found.) The variable @code{final_sequence} is null when not
8410 processing a sequence, otherwise it contains the @code{sequence} rtx
8414 @defmac REGISTER_PREFIX
8415 @defmacx LOCAL_LABEL_PREFIX
8416 @defmacx USER_LABEL_PREFIX
8417 @defmacx IMMEDIATE_PREFIX
8418 If defined, C string expressions to be used for the @samp{%R}, @samp{%L},
8419 @samp{%U}, and @samp{%I} options of @code{asm_fprintf} (see
8420 @file{final.c}). These are useful when a single @file{md} file must
8421 support multiple assembler formats. In that case, the various @file{tm.h}
8422 files can define these macros differently.
8425 @defmac ASM_FPRINTF_EXTENSIONS (@var{file}, @var{argptr}, @var{format})
8426 If defined this macro should expand to a series of @code{case}
8427 statements which will be parsed inside the @code{switch} statement of
8428 the @code{asm_fprintf} function. This allows targets to define extra
8429 printf formats which may useful when generating their assembler
8430 statements. Note that uppercase letters are reserved for future
8431 generic extensions to asm_fprintf, and so are not available to target
8432 specific code. The output file is given by the parameter @var{file}.
8433 The varargs input pointer is @var{argptr} and the rest of the format
8434 string, starting the character after the one that is being switched
8435 upon, is pointed to by @var{format}.
8438 @defmac ASSEMBLER_DIALECT
8439 If your target supports multiple dialects of assembler language (such as
8440 different opcodes), define this macro as a C expression that gives the
8441 numeric index of the assembler language dialect to use, with zero as the
8444 If this macro is defined, you may use constructs of the form
8446 @samp{@{option0|option1|option2@dots{}@}}
8449 in the output templates of patterns (@pxref{Output Template}) or in the
8450 first argument of @code{asm_fprintf}. This construct outputs
8451 @samp{option0}, @samp{option1}, @samp{option2}, etc., if the value of
8452 @code{ASSEMBLER_DIALECT} is zero, one, two, etc. Any special characters
8453 within these strings retain their usual meaning. If there are fewer
8454 alternatives within the braces than the value of
8455 @code{ASSEMBLER_DIALECT}, the construct outputs nothing.
8457 If you do not define this macro, the characters @samp{@{}, @samp{|} and
8458 @samp{@}} do not have any special meaning when used in templates or
8459 operands to @code{asm_fprintf}.
8461 Define the macros @code{REGISTER_PREFIX}, @code{LOCAL_LABEL_PREFIX},
8462 @code{USER_LABEL_PREFIX} and @code{IMMEDIATE_PREFIX} if you can express
8463 the variations in assembler language syntax with that mechanism. Define
8464 @code{ASSEMBLER_DIALECT} and use the @samp{@{option0|option1@}} syntax
8465 if the syntax variant are larger and involve such things as different
8466 opcodes or operand order.
8469 @defmac ASM_OUTPUT_REG_PUSH (@var{stream}, @var{regno})
8470 A C expression to output to @var{stream} some assembler code
8471 which will push hard register number @var{regno} onto the stack.
8472 The code need not be optimal, since this macro is used only when
8476 @defmac ASM_OUTPUT_REG_POP (@var{stream}, @var{regno})
8477 A C expression to output to @var{stream} some assembler code
8478 which will pop hard register number @var{regno} off of the stack.
8479 The code need not be optimal, since this macro is used only when
8483 @node Dispatch Tables
8484 @subsection Output of Dispatch Tables
8486 @c prevent bad page break with this line
8487 This concerns dispatch tables.
8489 @cindex dispatch table
8490 @defmac ASM_OUTPUT_ADDR_DIFF_ELT (@var{stream}, @var{body}, @var{value}, @var{rel})
8491 A C statement to output to the stdio stream @var{stream} an assembler
8492 pseudo-instruction to generate a difference between two labels.
8493 @var{value} and @var{rel} are the numbers of two internal labels. The
8494 definitions of these labels are output using
8495 @code{(*targetm.asm_out.internal_label)}, and they must be printed in the same
8496 way here. For example,
8499 fprintf (@var{stream}, "\t.word L%d-L%d\n",
8500 @var{value}, @var{rel})
8503 You must provide this macro on machines where the addresses in a
8504 dispatch table are relative to the table's own address. If defined, GCC
8505 will also use this macro on all machines when producing PIC@.
8506 @var{body} is the body of the @code{ADDR_DIFF_VEC}; it is provided so that the
8507 mode and flags can be read.
8510 @defmac ASM_OUTPUT_ADDR_VEC_ELT (@var{stream}, @var{value})
8511 This macro should be provided on machines where the addresses
8512 in a dispatch table are absolute.
8514 The definition should be a C statement to output to the stdio stream
8515 @var{stream} an assembler pseudo-instruction to generate a reference to
8516 a label. @var{value} is the number of an internal label whose
8517 definition is output using @code{(*targetm.asm_out.internal_label)}.
8521 fprintf (@var{stream}, "\t.word L%d\n", @var{value})
8525 @defmac ASM_OUTPUT_CASE_LABEL (@var{stream}, @var{prefix}, @var{num}, @var{table})
8526 Define this if the label before a jump-table needs to be output
8527 specially. The first three arguments are the same as for
8528 @code{(*targetm.asm_out.internal_label)}; the fourth argument is the
8529 jump-table which follows (a @code{jump_insn} containing an
8530 @code{addr_vec} or @code{addr_diff_vec}).
8532 This feature is used on system V to output a @code{swbeg} statement
8535 If this macro is not defined, these labels are output with
8536 @code{(*targetm.asm_out.internal_label)}.
8539 @defmac ASM_OUTPUT_CASE_END (@var{stream}, @var{num}, @var{table})
8540 Define this if something special must be output at the end of a
8541 jump-table. The definition should be a C statement to be executed
8542 after the assembler code for the table is written. It should write
8543 the appropriate code to stdio stream @var{stream}. The argument
8544 @var{table} is the jump-table insn, and @var{num} is the label-number
8545 of the preceding label.
8547 If this macro is not defined, nothing special is output at the end of
8551 @deftypefn {Target Hook} void TARGET_ASM_EMIT_UNWIND_LABEL (@var{stream}, @var{decl}, @var{for_eh}, @var{empty})
8552 This target hook emits a label at the beginning of each FDE@. It
8553 should be defined on targets where FDEs need special labels, and it
8554 should write the appropriate label, for the FDE associated with the
8555 function declaration @var{decl}, to the stdio stream @var{stream}.
8556 The third argument, @var{for_eh}, is a boolean: true if this is for an
8557 exception table. The fourth argument, @var{empty}, is a boolean:
8558 true if this is a placeholder label for an omitted FDE@.
8560 The default is that FDEs are not given nonlocal labels.
8563 @deftypefn {Target Hook} void TARGET_ASM_EMIT_EXCEPT_TABLE_LABEL (@var{stream})
8564 This target hook emits a label at the beginning of the exception table.
8565 It should be defined on targets where it is desirable for the table
8566 to be broken up according to function.
8568 The default is that no label is emitted.
8571 @deftypefn {Target Hook} void TARGET_UNWIND_EMIT (FILE * @var{stream}, rtx @var{insn})
8572 This target hook emits and assembly directives required to unwind the
8573 given instruction. This is only used when TARGET_UNWIND_INFO is set.
8576 @node Exception Region Output
8577 @subsection Assembler Commands for Exception Regions
8579 @c prevent bad page break with this line
8581 This describes commands marking the start and the end of an exception
8584 @defmac EH_FRAME_SECTION_NAME
8585 If defined, a C string constant for the name of the section containing
8586 exception handling frame unwind information. If not defined, GCC will
8587 provide a default definition if the target supports named sections.
8588 @file{crtstuff.c} uses this macro to switch to the appropriate section.
8590 You should define this symbol if your target supports DWARF 2 frame
8591 unwind information and the default definition does not work.
8594 @defmac EH_FRAME_IN_DATA_SECTION
8595 If defined, DWARF 2 frame unwind information will be placed in the
8596 data section even though the target supports named sections. This
8597 might be necessary, for instance, if the system linker does garbage
8598 collection and sections cannot be marked as not to be collected.
8600 Do not define this macro unless @code{TARGET_ASM_NAMED_SECTION} is
8604 @defmac EH_TABLES_CAN_BE_READ_ONLY
8605 Define this macro to 1 if your target is such that no frame unwind
8606 information encoding used with non-PIC code will ever require a
8607 runtime relocation, but the linker may not support merging read-only
8608 and read-write sections into a single read-write section.
8611 @defmac MASK_RETURN_ADDR
8612 An rtx used to mask the return address found via @code{RETURN_ADDR_RTX}, so
8613 that it does not contain any extraneous set bits in it.
8616 @defmac DWARF2_UNWIND_INFO
8617 Define this macro to 0 if your target supports DWARF 2 frame unwind
8618 information, but it does not yet work with exception handling.
8619 Otherwise, if your target supports this information (if it defines
8620 @samp{INCOMING_RETURN_ADDR_RTX} and either @samp{UNALIGNED_INT_ASM_OP}
8621 or @samp{OBJECT_FORMAT_ELF}), GCC will provide a default definition of 1.
8623 If @code{TARGET_UNWIND_INFO} is defined, the target specific unwinder
8624 will be used in all cases. Defining this macro will enable the generation
8625 of DWARF 2 frame debugging information.
8627 If @code{TARGET_UNWIND_INFO} is not defined, and this macro is defined to 1,
8628 the DWARF 2 unwinder will be the default exception handling mechanism;
8629 otherwise, the @code{setjmp}/@code{longjmp}-based scheme will be used by
8633 @defmac TARGET_UNWIND_INFO
8634 Define this macro if your target has ABI specified unwind tables. Usually
8635 these will be output by @code{TARGET_UNWIND_EMIT}.
8638 @deftypevr {Target Hook} bool TARGET_UNWIND_TABLES_DEFAULT
8639 This variable should be set to @code{true} if the target ABI requires unwinding
8640 tables even when exceptions are not used.
8643 @defmac MUST_USE_SJLJ_EXCEPTIONS
8644 This macro need only be defined if @code{DWARF2_UNWIND_INFO} is
8645 runtime-variable. In that case, @file{except.h} cannot correctly
8646 determine the corresponding definition of @code{MUST_USE_SJLJ_EXCEPTIONS},
8647 so the target must provide it directly.
8650 @defmac DONT_USE_BUILTIN_SETJMP
8651 Define this macro to 1 if the @code{setjmp}/@code{longjmp}-based scheme
8652 should use the @code{setjmp}/@code{longjmp} functions from the C library
8653 instead of the @code{__builtin_setjmp}/@code{__builtin_longjmp} machinery.
8656 @defmac DWARF_CIE_DATA_ALIGNMENT
8657 This macro need only be defined if the target might save registers in the
8658 function prologue at an offset to the stack pointer that is not aligned to
8659 @code{UNITS_PER_WORD}. The definition should be the negative minimum
8660 alignment if @code{STACK_GROWS_DOWNWARD} is defined, and the positive
8661 minimum alignment otherwise. @xref{SDB and DWARF}. Only applicable if
8662 the target supports DWARF 2 frame unwind information.
8665 @deftypevr {Target Hook} bool TARGET_TERMINATE_DW2_EH_FRAME_INFO
8666 Contains the value true if the target should add a zero word onto the
8667 end of a Dwarf-2 frame info section when used for exception handling.
8668 Default value is false if @code{EH_FRAME_SECTION_NAME} is defined, and
8672 @deftypefn {Target Hook} rtx TARGET_DWARF_REGISTER_SPAN (rtx @var{reg})
8673 Given a register, this hook should return a parallel of registers to
8674 represent where to find the register pieces. Define this hook if the
8675 register and its mode are represented in Dwarf in non-contiguous
8676 locations, or if the register should be represented in more than one
8677 register in Dwarf. Otherwise, this hook should return @code{NULL_RTX}.
8678 If not defined, the default is to return @code{NULL_RTX}.
8681 @deftypefn {Target Hook} void TARGET_INIT_DWARF_REG_SIZES_EXTRA (tree @var{address})
8682 If some registers are represented in Dwarf-2 unwind information in
8683 multiple pieces, define this hook to fill in information about the
8684 sizes of those pieces in the table used by the unwinder at runtime.
8685 It will be called by @code{expand_builtin_init_dwarf_reg_sizes} after
8686 filling in a single size corresponding to each hard register;
8687 @var{address} is the address of the table.
8690 @deftypefn {Target Hook} bool TARGET_ASM_TTYPE (rtx @var{sym})
8691 This hook is used to output a reference from a frame unwinding table to
8692 the type_info object identified by @var{sym}. It should return @code{true}
8693 if the reference was output. Returning @code{false} will cause the
8694 reference to be output using the normal Dwarf2 routines.
8697 @deftypefn {Target Hook} bool TARGET_ARM_EABI_UNWINDER
8698 This hook should be set to @code{true} on targets that use an ARM EABI
8699 based unwinding library, and @code{false} on other targets. This effects
8700 the format of unwinding tables, and how the unwinder in entered after
8701 running a cleanup. The default is @code{false}.
8704 @node Alignment Output
8705 @subsection Assembler Commands for Alignment
8707 @c prevent bad page break with this line
8708 This describes commands for alignment.
8710 @defmac JUMP_ALIGN (@var{label})
8711 The alignment (log base 2) to put in front of @var{label}, which is
8712 a common destination of jumps and has no fallthru incoming edge.
8714 This macro need not be defined if you don't want any special alignment
8715 to be done at such a time. Most machine descriptions do not currently
8718 Unless it's necessary to inspect the @var{label} parameter, it is better
8719 to set the variable @var{align_jumps} in the target's
8720 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8721 selection in @var{align_jumps} in a @code{JUMP_ALIGN} implementation.
8724 @defmac LABEL_ALIGN_AFTER_BARRIER (@var{label})
8725 The alignment (log base 2) to put in front of @var{label}, which follows
8728 This macro need not be defined if you don't want any special alignment
8729 to be done at such a time. Most machine descriptions do not currently
8733 @defmac LABEL_ALIGN_AFTER_BARRIER_MAX_SKIP
8734 The maximum number of bytes to skip when applying
8735 @code{LABEL_ALIGN_AFTER_BARRIER}. This works only if
8736 @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8739 @defmac LOOP_ALIGN (@var{label})
8740 The alignment (log base 2) to put in front of @var{label}, which follows
8741 a @code{NOTE_INSN_LOOP_BEG} note.
8743 This macro need not be defined if you don't want any special alignment
8744 to be done at such a time. Most machine descriptions do not currently
8747 Unless it's necessary to inspect the @var{label} parameter, it is better
8748 to set the variable @code{align_loops} in the target's
8749 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8750 selection in @code{align_loops} in a @code{LOOP_ALIGN} implementation.
8753 @defmac LOOP_ALIGN_MAX_SKIP
8754 The maximum number of bytes to skip when applying @code{LOOP_ALIGN}.
8755 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8758 @defmac LABEL_ALIGN (@var{label})
8759 The alignment (log base 2) to put in front of @var{label}.
8760 If @code{LABEL_ALIGN_AFTER_BARRIER} / @code{LOOP_ALIGN} specify a different alignment,
8761 the maximum of the specified values is used.
8763 Unless it's necessary to inspect the @var{label} parameter, it is better
8764 to set the variable @code{align_labels} in the target's
8765 @code{OVERRIDE_OPTIONS}. Otherwise, you should try to honor the user's
8766 selection in @code{align_labels} in a @code{LABEL_ALIGN} implementation.
8769 @defmac LABEL_ALIGN_MAX_SKIP
8770 The maximum number of bytes to skip when applying @code{LABEL_ALIGN}.
8771 This works only if @code{ASM_OUTPUT_MAX_SKIP_ALIGN} is defined.
8774 @defmac ASM_OUTPUT_SKIP (@var{stream}, @var{nbytes})
8775 A C statement to output to the stdio stream @var{stream} an assembler
8776 instruction to advance the location counter by @var{nbytes} bytes.
8777 Those bytes should be zero when loaded. @var{nbytes} will be a C
8778 expression of type @code{unsigned HOST_WIDE_INT}.
8781 @defmac ASM_NO_SKIP_IN_TEXT
8782 Define this macro if @code{ASM_OUTPUT_SKIP} should not be used in the
8783 text section because it fails to put zeros in the bytes that are skipped.
8784 This is true on many Unix systems, where the pseudo--op to skip bytes
8785 produces no-op instructions rather than zeros when used in the text
8789 @defmac ASM_OUTPUT_ALIGN (@var{stream}, @var{power})
8790 A C statement to output to the stdio stream @var{stream} an assembler
8791 command to advance the location counter to a multiple of 2 to the
8792 @var{power} bytes. @var{power} will be a C expression of type @code{int}.
8795 @defmac ASM_OUTPUT_ALIGN_WITH_NOP (@var{stream}, @var{power})
8796 Like @code{ASM_OUTPUT_ALIGN}, except that the ``nop'' instruction is used
8797 for padding, if necessary.
8800 @defmac ASM_OUTPUT_MAX_SKIP_ALIGN (@var{stream}, @var{power}, @var{max_skip})
8801 A C statement to output to the stdio stream @var{stream} an assembler
8802 command to advance the location counter to a multiple of 2 to the
8803 @var{power} bytes, but only if @var{max_skip} or fewer bytes are needed to
8804 satisfy the alignment request. @var{power} and @var{max_skip} will be
8805 a C expression of type @code{int}.
8809 @node Debugging Info
8810 @section Controlling Debugging Information Format
8812 @c prevent bad page break with this line
8813 This describes how to specify debugging information.
8816 * All Debuggers:: Macros that affect all debugging formats uniformly.
8817 * DBX Options:: Macros enabling specific options in DBX format.
8818 * DBX Hooks:: Hook macros for varying DBX format.
8819 * File Names and DBX:: Macros controlling output of file names in DBX format.
8820 * SDB and DWARF:: Macros for SDB (COFF) and DWARF formats.
8821 * VMS Debug:: Macros for VMS debug format.
8825 @subsection Macros Affecting All Debugging Formats
8827 @c prevent bad page break with this line
8828 These macros affect all debugging formats.
8830 @defmac DBX_REGISTER_NUMBER (@var{regno})
8831 A C expression that returns the DBX register number for the compiler
8832 register number @var{regno}. In the default macro provided, the value
8833 of this expression will be @var{regno} itself. But sometimes there are
8834 some registers that the compiler knows about and DBX does not, or vice
8835 versa. In such cases, some register may need to have one number in the
8836 compiler and another for DBX@.
8838 If two registers have consecutive numbers inside GCC, and they can be
8839 used as a pair to hold a multiword value, then they @emph{must} have
8840 consecutive numbers after renumbering with @code{DBX_REGISTER_NUMBER}.
8841 Otherwise, debuggers will be unable to access such a pair, because they
8842 expect register pairs to be consecutive in their own numbering scheme.
8844 If you find yourself defining @code{DBX_REGISTER_NUMBER} in way that
8845 does not preserve register pairs, then what you must do instead is
8846 redefine the actual register numbering scheme.
8849 @defmac DEBUGGER_AUTO_OFFSET (@var{x})
8850 A C expression that returns the integer offset value for an automatic
8851 variable having address @var{x} (an RTL expression). The default
8852 computation assumes that @var{x} is based on the frame-pointer and
8853 gives the offset from the frame-pointer. This is required for targets
8854 that produce debugging output for DBX or COFF-style debugging output
8855 for SDB and allow the frame-pointer to be eliminated when the
8856 @option{-g} options is used.
8859 @defmac DEBUGGER_ARG_OFFSET (@var{offset}, @var{x})
8860 A C expression that returns the integer offset value for an argument
8861 having address @var{x} (an RTL expression). The nominal offset is
8865 @defmac PREFERRED_DEBUGGING_TYPE
8866 A C expression that returns the type of debugging output GCC should
8867 produce when the user specifies just @option{-g}. Define
8868 this if you have arranged for GCC to support more than one format of
8869 debugging output. Currently, the allowable values are @code{DBX_DEBUG},
8870 @code{SDB_DEBUG}, @code{DWARF_DEBUG}, @code{DWARF2_DEBUG},
8871 @code{XCOFF_DEBUG}, @code{VMS_DEBUG}, and @code{VMS_AND_DWARF2_DEBUG}.
8873 When the user specifies @option{-ggdb}, GCC normally also uses the
8874 value of this macro to select the debugging output format, but with two
8875 exceptions. If @code{DWARF2_DEBUGGING_INFO} is defined, GCC uses the
8876 value @code{DWARF2_DEBUG}. Otherwise, if @code{DBX_DEBUGGING_INFO} is
8877 defined, GCC uses @code{DBX_DEBUG}.
8879 The value of this macro only affects the default debugging output; the
8880 user can always get a specific type of output by using @option{-gstabs},
8881 @option{-gcoff}, @option{-gdwarf-2}, @option{-gxcoff}, or @option{-gvms}.
8885 @subsection Specific Options for DBX Output
8887 @c prevent bad page break with this line
8888 These are specific options for DBX output.
8890 @defmac DBX_DEBUGGING_INFO
8891 Define this macro if GCC should produce debugging output for DBX
8892 in response to the @option{-g} option.
8895 @defmac XCOFF_DEBUGGING_INFO
8896 Define this macro if GCC should produce XCOFF format debugging output
8897 in response to the @option{-g} option. This is a variant of DBX format.
8900 @defmac DEFAULT_GDB_EXTENSIONS
8901 Define this macro to control whether GCC should by default generate
8902 GDB's extended version of DBX debugging information (assuming DBX-format
8903 debugging information is enabled at all). If you don't define the
8904 macro, the default is 1: always generate the extended information
8905 if there is any occasion to.
8908 @defmac DEBUG_SYMS_TEXT
8909 Define this macro if all @code{.stabs} commands should be output while
8910 in the text section.
8913 @defmac ASM_STABS_OP
8914 A C string constant, including spacing, naming the assembler pseudo op to
8915 use instead of @code{"\t.stabs\t"} to define an ordinary debugging symbol.
8916 If you don't define this macro, @code{"\t.stabs\t"} is used. This macro
8917 applies only to DBX debugging information format.
8920 @defmac ASM_STABD_OP
8921 A C string constant, including spacing, naming the assembler pseudo op to
8922 use instead of @code{"\t.stabd\t"} to define a debugging symbol whose
8923 value is the current location. If you don't define this macro,
8924 @code{"\t.stabd\t"} is used. This macro applies only to DBX debugging
8928 @defmac ASM_STABN_OP
8929 A C string constant, including spacing, naming the assembler pseudo op to
8930 use instead of @code{"\t.stabn\t"} to define a debugging symbol with no
8931 name. If you don't define this macro, @code{"\t.stabn\t"} is used. This
8932 macro applies only to DBX debugging information format.
8935 @defmac DBX_NO_XREFS
8936 Define this macro if DBX on your system does not support the construct
8937 @samp{xs@var{tagname}}. On some systems, this construct is used to
8938 describe a forward reference to a structure named @var{tagname}.
8939 On other systems, this construct is not supported at all.
8942 @defmac DBX_CONTIN_LENGTH
8943 A symbol name in DBX-format debugging information is normally
8944 continued (split into two separate @code{.stabs} directives) when it
8945 exceeds a certain length (by default, 80 characters). On some
8946 operating systems, DBX requires this splitting; on others, splitting
8947 must not be done. You can inhibit splitting by defining this macro
8948 with the value zero. You can override the default splitting-length by
8949 defining this macro as an expression for the length you desire.
8952 @defmac DBX_CONTIN_CHAR
8953 Normally continuation is indicated by adding a @samp{\} character to
8954 the end of a @code{.stabs} string when a continuation follows. To use
8955 a different character instead, define this macro as a character
8956 constant for the character you want to use. Do not define this macro
8957 if backslash is correct for your system.
8960 @defmac DBX_STATIC_STAB_DATA_SECTION
8961 Define this macro if it is necessary to go to the data section before
8962 outputting the @samp{.stabs} pseudo-op for a non-global static
8966 @defmac DBX_TYPE_DECL_STABS_CODE
8967 The value to use in the ``code'' field of the @code{.stabs} directive
8968 for a typedef. The default is @code{N_LSYM}.
8971 @defmac DBX_STATIC_CONST_VAR_CODE
8972 The value to use in the ``code'' field of the @code{.stabs} directive
8973 for a static variable located in the text section. DBX format does not
8974 provide any ``right'' way to do this. The default is @code{N_FUN}.
8977 @defmac DBX_REGPARM_STABS_CODE
8978 The value to use in the ``code'' field of the @code{.stabs} directive
8979 for a parameter passed in registers. DBX format does not provide any
8980 ``right'' way to do this. The default is @code{N_RSYM}.
8983 @defmac DBX_REGPARM_STABS_LETTER
8984 The letter to use in DBX symbol data to identify a symbol as a parameter
8985 passed in registers. DBX format does not customarily provide any way to
8986 do this. The default is @code{'P'}.
8989 @defmac DBX_FUNCTION_FIRST
8990 Define this macro if the DBX information for a function and its
8991 arguments should precede the assembler code for the function. Normally,
8992 in DBX format, the debugging information entirely follows the assembler
8996 @defmac DBX_BLOCKS_FUNCTION_RELATIVE
8997 Define this macro, with value 1, if the value of a symbol describing
8998 the scope of a block (@code{N_LBRAC} or @code{N_RBRAC}) should be
8999 relative to the start of the enclosing function. Normally, GCC uses
9000 an absolute address.
9003 @defmac DBX_LINES_FUNCTION_RELATIVE
9004 Define this macro, with value 1, if the value of a symbol indicating
9005 the current line number (@code{N_SLINE}) should be relative to the
9006 start of the enclosing function. Normally, GCC uses an absolute address.
9009 @defmac DBX_USE_BINCL
9010 Define this macro if GCC should generate @code{N_BINCL} and
9011 @code{N_EINCL} stabs for included header files, as on Sun systems. This
9012 macro also directs GCC to output a type number as a pair of a file
9013 number and a type number within the file. Normally, GCC does not
9014 generate @code{N_BINCL} or @code{N_EINCL} stabs, and it outputs a single
9015 number for a type number.
9019 @subsection Open-Ended Hooks for DBX Format
9021 @c prevent bad page break with this line
9022 These are hooks for DBX format.
9024 @defmac DBX_OUTPUT_LBRAC (@var{stream}, @var{name})
9025 Define this macro to say how to output to @var{stream} the debugging
9026 information for the start of a scope level for variable names. The
9027 argument @var{name} is the name of an assembler symbol (for use with
9028 @code{assemble_name}) whose value is the address where the scope begins.
9031 @defmac DBX_OUTPUT_RBRAC (@var{stream}, @var{name})
9032 Like @code{DBX_OUTPUT_LBRAC}, but for the end of a scope level.
9035 @defmac DBX_OUTPUT_NFUN (@var{stream}, @var{lscope_label}, @var{decl})
9036 Define this macro if the target machine requires special handling to
9037 output an @code{N_FUN} entry for the function @var{decl}.
9040 @defmac DBX_OUTPUT_SOURCE_LINE (@var{stream}, @var{line}, @var{counter})
9041 A C statement to output DBX debugging information before code for line
9042 number @var{line} of the current source file to the stdio stream
9043 @var{stream}. @var{counter} is the number of time the macro was
9044 invoked, including the current invocation; it is intended to generate
9045 unique labels in the assembly output.
9047 This macro should not be defined if the default output is correct, or
9048 if it can be made correct by defining @code{DBX_LINES_FUNCTION_RELATIVE}.
9051 @defmac NO_DBX_FUNCTION_END
9052 Some stabs encapsulation formats (in particular ECOFF), cannot handle the
9053 @code{.stabs "",N_FUN,,0,0,Lscope-function-1} gdb dbx extension construct.
9054 On those machines, define this macro to turn this feature off without
9055 disturbing the rest of the gdb extensions.
9058 @defmac NO_DBX_BNSYM_ENSYM
9059 Some assemblers cannot handle the @code{.stabd BNSYM/ENSYM,0,0} gdb dbx
9060 extension construct. On those machines, define this macro to turn this
9061 feature off without disturbing the rest of the gdb extensions.
9064 @node File Names and DBX
9065 @subsection File Names in DBX Format
9067 @c prevent bad page break with this line
9068 This describes file names in DBX format.
9070 @defmac DBX_OUTPUT_MAIN_SOURCE_FILENAME (@var{stream}, @var{name})
9071 A C statement to output DBX debugging information to the stdio stream
9072 @var{stream}, which indicates that file @var{name} is the main source
9073 file---the file specified as the input file for compilation.
9074 This macro is called only once, at the beginning of compilation.
9076 This macro need not be defined if the standard form of output
9077 for DBX debugging information is appropriate.
9079 It may be necessary to refer to a label equal to the beginning of the
9080 text section. You can use @samp{assemble_name (stream, ltext_label_name)}
9081 to do so. If you do this, you must also set the variable
9082 @var{used_ltext_label_name} to @code{true}.
9085 @defmac NO_DBX_MAIN_SOURCE_DIRECTORY
9086 Define this macro, with value 1, if GCC should not emit an indication
9087 of the current directory for compilation and current source language at
9088 the beginning of the file.
9091 @defmac NO_DBX_GCC_MARKER
9092 Define this macro, with value 1, if GCC should not emit an indication
9093 that this object file was compiled by GCC@. The default is to emit
9094 an @code{N_OPT} stab at the beginning of every source file, with
9095 @samp{gcc2_compiled.} for the string and value 0.
9098 @defmac DBX_OUTPUT_MAIN_SOURCE_FILE_END (@var{stream}, @var{name})
9099 A C statement to output DBX debugging information at the end of
9100 compilation of the main source file @var{name}. Output should be
9101 written to the stdio stream @var{stream}.
9103 If you don't define this macro, nothing special is output at the end
9104 of compilation, which is correct for most machines.
9107 @defmac DBX_OUTPUT_NULL_N_SO_AT_MAIN_SOURCE_FILE_END
9108 Define this macro @emph{instead of} defining
9109 @code{DBX_OUTPUT_MAIN_SOURCE_FILE_END}, if what needs to be output at
9110 the end of compilation is an @code{N_SO} stab with an empty string,
9111 whose value is the highest absolute text address in the file.
9116 @subsection Macros for SDB and DWARF Output
9118 @c prevent bad page break with this line
9119 Here are macros for SDB and DWARF output.
9121 @defmac SDB_DEBUGGING_INFO
9122 Define this macro if GCC should produce COFF-style debugging output
9123 for SDB in response to the @option{-g} option.
9126 @defmac DWARF2_DEBUGGING_INFO
9127 Define this macro if GCC should produce dwarf version 2 format
9128 debugging output in response to the @option{-g} option.
9130 @deftypefn {Target Hook} int TARGET_DWARF_CALLING_CONVENTION (tree @var{function})
9131 Define this to enable the dwarf attribute @code{DW_AT_calling_convention} to
9132 be emitted for each function. Instead of an integer return the enum
9133 value for the @code{DW_CC_} tag.
9136 To support optional call frame debugging information, you must also
9137 define @code{INCOMING_RETURN_ADDR_RTX} and either set
9138 @code{RTX_FRAME_RELATED_P} on the prologue insns if you use RTL for the
9139 prologue, or call @code{dwarf2out_def_cfa} and @code{dwarf2out_reg_save}
9140 as appropriate from @code{TARGET_ASM_FUNCTION_PROLOGUE} if you don't.
9143 @defmac DWARF2_FRAME_INFO
9144 Define this macro to a nonzero value if GCC should always output
9145 Dwarf 2 frame information. If @code{DWARF2_UNWIND_INFO}
9146 (@pxref{Exception Region Output} is nonzero, GCC will output this
9147 information not matter how you define @code{DWARF2_FRAME_INFO}.
9150 @defmac DWARF2_ASM_LINE_DEBUG_INFO
9151 Define this macro to be a nonzero value if the assembler can generate Dwarf 2
9152 line debug info sections. This will result in much more compact line number
9153 tables, and hence is desirable if it works.
9156 @defmac ASM_OUTPUT_DWARF_DELTA (@var{stream}, @var{size}, @var{label1}, @var{label2})
9157 A C statement to issue assembly directives that create a difference
9158 @var{lab1} minus @var{lab2}, using an integer of the given @var{size}.
9161 @defmac ASM_OUTPUT_DWARF_OFFSET (@var{stream}, @var{size}, @var{label}, @var{section})
9162 A C statement to issue assembly directives that create a
9163 section-relative reference to the given @var{label}, using an integer of the
9164 given @var{size}. The label is known to be defined in the given @var{section}.
9167 @defmac ASM_OUTPUT_DWARF_PCREL (@var{stream}, @var{size}, @var{label})
9168 A C statement to issue assembly directives that create a self-relative
9169 reference to the given @var{label}, using an integer of the given @var{size}.
9172 @defmac ASM_OUTPUT_DWARF_TABLE_REF (@var{label})
9173 A C statement to issue assembly directives that create a reference to
9174 the DWARF table identifier @var{label} from the current section. This
9175 is used on some systems to avoid garbage collecting a DWARF table which
9176 is referenced by a function.
9179 @deftypefn {Target Hook} void TARGET_ASM_OUTPUT_DWARF_DTPREL (FILE *@var{FILE}, int @var{size}, rtx @var{x})
9180 If defined, this target hook is a function which outputs a DTP-relative
9181 reference to the given TLS symbol of the specified size.
9184 @defmac PUT_SDB_@dots{}
9185 Define these macros to override the assembler syntax for the special
9186 SDB assembler directives. See @file{sdbout.c} for a list of these
9187 macros and their arguments. If the standard syntax is used, you need
9188 not define them yourself.
9192 Some assemblers do not support a semicolon as a delimiter, even between
9193 SDB assembler directives. In that case, define this macro to be the
9194 delimiter to use (usually @samp{\n}). It is not necessary to define
9195 a new set of @code{PUT_SDB_@var{op}} macros if this is the only change
9199 @defmac SDB_ALLOW_UNKNOWN_REFERENCES
9200 Define this macro to allow references to unknown structure,
9201 union, or enumeration tags to be emitted. Standard COFF does not
9202 allow handling of unknown references, MIPS ECOFF has support for
9206 @defmac SDB_ALLOW_FORWARD_REFERENCES
9207 Define this macro to allow references to structure, union, or
9208 enumeration tags that have not yet been seen to be handled. Some
9209 assemblers choke if forward tags are used, while some require it.
9212 @defmac SDB_OUTPUT_SOURCE_LINE (@var{stream}, @var{line})
9213 A C statement to output SDB debugging information before code for line
9214 number @var{line} of the current source file to the stdio stream
9215 @var{stream}. The default is to emit an @code{.ln} directive.
9220 @subsection Macros for VMS Debug Format
9222 @c prevent bad page break with this line
9223 Here are macros for VMS debug format.
9225 @defmac VMS_DEBUGGING_INFO
9226 Define this macro if GCC should produce debugging output for VMS
9227 in response to the @option{-g} option. The default behavior for VMS
9228 is to generate minimal debug info for a traceback in the absence of
9229 @option{-g} unless explicitly overridden with @option{-g0}. This
9230 behavior is controlled by @code{OPTIMIZATION_OPTIONS} and
9231 @code{OVERRIDE_OPTIONS}.
9234 @node Floating Point
9235 @section Cross Compilation and Floating Point
9236 @cindex cross compilation and floating point
9237 @cindex floating point and cross compilation
9239 While all modern machines use twos-complement representation for integers,
9240 there are a variety of representations for floating point numbers. This
9241 means that in a cross-compiler the representation of floating point numbers
9242 in the compiled program may be different from that used in the machine
9243 doing the compilation.
9245 Because different representation systems may offer different amounts of
9246 range and precision, all floating point constants must be represented in
9247 the target machine's format. Therefore, the cross compiler cannot
9248 safely use the host machine's floating point arithmetic; it must emulate
9249 the target's arithmetic. To ensure consistency, GCC always uses
9250 emulation to work with floating point values, even when the host and
9251 target floating point formats are identical.
9253 The following macros are provided by @file{real.h} for the compiler to
9254 use. All parts of the compiler which generate or optimize
9255 floating-point calculations must use these macros. They may evaluate
9256 their operands more than once, so operands must not have side effects.
9258 @defmac REAL_VALUE_TYPE
9259 The C data type to be used to hold a floating point value in the target
9260 machine's format. Typically this is a @code{struct} containing an
9261 array of @code{HOST_WIDE_INT}, but all code should treat it as an opaque
9265 @deftypefn Macro int REAL_VALUES_EQUAL (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9266 Compares for equality the two values, @var{x} and @var{y}. If the target
9267 floating point format supports negative zeroes and/or NaNs,
9268 @samp{REAL_VALUES_EQUAL (-0.0, 0.0)} is true, and
9269 @samp{REAL_VALUES_EQUAL (NaN, NaN)} is false.
9272 @deftypefn Macro int REAL_VALUES_LESS (REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9273 Tests whether @var{x} is less than @var{y}.
9276 @deftypefn Macro HOST_WIDE_INT REAL_VALUE_FIX (REAL_VALUE_TYPE @var{x})
9277 Truncates @var{x} to a signed integer, rounding toward zero.
9280 @deftypefn Macro {unsigned HOST_WIDE_INT} REAL_VALUE_UNSIGNED_FIX (REAL_VALUE_TYPE @var{x})
9281 Truncates @var{x} to an unsigned integer, rounding toward zero. If
9282 @var{x} is negative, returns zero.
9285 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ATOF (const char *@var{string}, enum machine_mode @var{mode})
9286 Converts @var{string} into a floating point number in the target machine's
9287 representation for mode @var{mode}. This routine can handle both
9288 decimal and hexadecimal floating point constants, using the syntax
9289 defined by the C language for both.
9292 @deftypefn Macro int REAL_VALUE_NEGATIVE (REAL_VALUE_TYPE @var{x})
9293 Returns 1 if @var{x} is negative (including negative zero), 0 otherwise.
9296 @deftypefn Macro int REAL_VALUE_ISINF (REAL_VALUE_TYPE @var{x})
9297 Determines whether @var{x} represents infinity (positive or negative).
9300 @deftypefn Macro int REAL_VALUE_ISNAN (REAL_VALUE_TYPE @var{x})
9301 Determines whether @var{x} represents a ``NaN'' (not-a-number).
9304 @deftypefn Macro void REAL_ARITHMETIC (REAL_VALUE_TYPE @var{output}, enum tree_code @var{code}, REAL_VALUE_TYPE @var{x}, REAL_VALUE_TYPE @var{y})
9305 Calculates an arithmetic operation on the two floating point values
9306 @var{x} and @var{y}, storing the result in @var{output} (which must be a
9309 The operation to be performed is specified by @var{code}. Only the
9310 following codes are supported: @code{PLUS_EXPR}, @code{MINUS_EXPR},
9311 @code{MULT_EXPR}, @code{RDIV_EXPR}, @code{MAX_EXPR}, @code{MIN_EXPR}.
9313 If @code{REAL_ARITHMETIC} is asked to evaluate division by zero and the
9314 target's floating point format cannot represent infinity, it will call
9315 @code{abort}. Callers should check for this situation first, using
9316 @code{MODE_HAS_INFINITIES}. @xref{Storage Layout}.
9319 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_NEGATE (REAL_VALUE_TYPE @var{x})
9320 Returns the negative of the floating point value @var{x}.
9323 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_ABS (REAL_VALUE_TYPE @var{x})
9324 Returns the absolute value of @var{x}.
9327 @deftypefn Macro REAL_VALUE_TYPE REAL_VALUE_TRUNCATE (REAL_VALUE_TYPE @var{mode}, enum machine_mode @var{x})
9328 Truncates the floating point value @var{x} to fit in @var{mode}. The
9329 return value is still a full-size @code{REAL_VALUE_TYPE}, but it has an
9330 appropriate bit pattern to be output as a floating constant whose
9331 precision accords with mode @var{mode}.
9334 @deftypefn Macro void REAL_VALUE_TO_INT (HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, REAL_VALUE_TYPE @var{x})
9335 Converts a floating point value @var{x} into a double-precision integer
9336 which is then stored into @var{low} and @var{high}. If the value is not
9337 integral, it is truncated.
9340 @deftypefn Macro void REAL_VALUE_FROM_INT (REAL_VALUE_TYPE @var{x}, HOST_WIDE_INT @var{low}, HOST_WIDE_INT @var{high}, enum machine_mode @var{mode})
9341 Converts a double-precision integer found in @var{low} and @var{high},
9342 into a floating point value which is then stored into @var{x}. The
9343 value is truncated to fit in mode @var{mode}.
9346 @node Mode Switching
9347 @section Mode Switching Instructions
9348 @cindex mode switching
9349 The following macros control mode switching optimizations:
9351 @defmac OPTIMIZE_MODE_SWITCHING (@var{entity})
9352 Define this macro if the port needs extra instructions inserted for mode
9353 switching in an optimizing compilation.
9355 For an example, the SH4 can perform both single and double precision
9356 floating point operations, but to perform a single precision operation,
9357 the FPSCR PR bit has to be cleared, while for a double precision
9358 operation, this bit has to be set. Changing the PR bit requires a general
9359 purpose register as a scratch register, hence these FPSCR sets have to
9360 be inserted before reload, i.e.@: you can't put this into instruction emitting
9361 or @code{TARGET_MACHINE_DEPENDENT_REORG}.
9363 You can have multiple entities that are mode-switched, and select at run time
9364 which entities actually need it. @code{OPTIMIZE_MODE_SWITCHING} should
9365 return nonzero for any @var{entity} that needs mode-switching.
9366 If you define this macro, you also have to define
9367 @code{NUM_MODES_FOR_MODE_SWITCHING}, @code{MODE_NEEDED},
9368 @code{MODE_PRIORITY_TO_MODE} and @code{EMIT_MODE_SET}.
9369 @code{MODE_AFTER}, @code{MODE_ENTRY}, and @code{MODE_EXIT}
9373 @defmac NUM_MODES_FOR_MODE_SWITCHING
9374 If you define @code{OPTIMIZE_MODE_SWITCHING}, you have to define this as
9375 initializer for an array of integers. Each initializer element
9376 N refers to an entity that needs mode switching, and specifies the number
9377 of different modes that might need to be set for this entity.
9378 The position of the initializer in the initializer---starting counting at
9379 zero---determines the integer that is used to refer to the mode-switched
9381 In macros that take mode arguments / yield a mode result, modes are
9382 represented as numbers 0 @dots{} N @minus{} 1. N is used to specify that no mode
9383 switch is needed / supplied.
9386 @defmac MODE_NEEDED (@var{entity}, @var{insn})
9387 @var{entity} is an integer specifying a mode-switched entity. If
9388 @code{OPTIMIZE_MODE_SWITCHING} is defined, you must define this macro to
9389 return an integer value not larger than the corresponding element in
9390 @code{NUM_MODES_FOR_MODE_SWITCHING}, to denote the mode that @var{entity} must
9391 be switched into prior to the execution of @var{insn}.
9394 @defmac MODE_AFTER (@var{mode}, @var{insn})
9395 If this macro is defined, it is evaluated for every @var{insn} during
9396 mode switching. It determines the mode that an insn results in (if
9397 different from the incoming mode).
9400 @defmac MODE_ENTRY (@var{entity})
9401 If this macro is defined, it is evaluated for every @var{entity} that needs
9402 mode switching. It should evaluate to an integer, which is a mode that
9403 @var{entity} is assumed to be switched to at function entry. If @code{MODE_ENTRY}
9404 is defined then @code{MODE_EXIT} must be defined.
9407 @defmac MODE_EXIT (@var{entity})
9408 If this macro is defined, it is evaluated for every @var{entity} that needs
9409 mode switching. It should evaluate to an integer, which is a mode that
9410 @var{entity} is assumed to be switched to at function exit. If @code{MODE_EXIT}
9411 is defined then @code{MODE_ENTRY} must be defined.
9414 @defmac MODE_PRIORITY_TO_MODE (@var{entity}, @var{n})
9415 This macro specifies the order in which modes for @var{entity} are processed.
9416 0 is the highest priority, @code{NUM_MODES_FOR_MODE_SWITCHING[@var{entity}] - 1} the
9417 lowest. The value of the macro should be an integer designating a mode
9418 for @var{entity}. For any fixed @var{entity}, @code{mode_priority_to_mode}
9419 (@var{entity}, @var{n}) shall be a bijection in 0 @dots{}
9420 @code{num_modes_for_mode_switching[@var{entity}] - 1}.
9423 @defmac EMIT_MODE_SET (@var{entity}, @var{mode}, @var{hard_regs_live})
9424 Generate one or more insns to set @var{entity} to @var{mode}.
9425 @var{hard_reg_live} is the set of hard registers live at the point where
9426 the insn(s) are to be inserted.
9429 @node Target Attributes
9430 @section Defining target-specific uses of @code{__attribute__}
9431 @cindex target attributes
9432 @cindex machine attributes
9433 @cindex attributes, target-specific
9435 Target-specific attributes may be defined for functions, data and types.
9436 These are described using the following target hooks; they also need to
9437 be documented in @file{extend.texi}.
9439 @deftypevr {Target Hook} {const struct attribute_spec *} TARGET_ATTRIBUTE_TABLE
9440 If defined, this target hook points to an array of @samp{struct
9441 attribute_spec} (defined in @file{tree.h}) specifying the machine
9442 specific attributes for this target and some of the restrictions on the
9443 entities to which these attributes are applied and the arguments they
9447 @deftypefn {Target Hook} int TARGET_COMP_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9448 If defined, this target hook is a function which returns zero if the attributes on
9449 @var{type1} and @var{type2} are incompatible, one if they are compatible,
9450 and two if they are nearly compatible (which causes a warning to be
9451 generated). If this is not defined, machine-specific attributes are
9452 supposed always to be compatible.
9455 @deftypefn {Target Hook} void TARGET_SET_DEFAULT_TYPE_ATTRIBUTES (tree @var{type})
9456 If defined, this target hook is a function which assigns default attributes to
9457 newly defined @var{type}.
9460 @deftypefn {Target Hook} tree TARGET_MERGE_TYPE_ATTRIBUTES (tree @var{type1}, tree @var{type2})
9461 Define this target hook if the merging of type attributes needs special
9462 handling. If defined, the result is a list of the combined
9463 @code{TYPE_ATTRIBUTES} of @var{type1} and @var{type2}. It is assumed
9464 that @code{comptypes} has already been called and returned 1. This
9465 function may call @code{merge_attributes} to handle machine-independent
9469 @deftypefn {Target Hook} tree TARGET_MERGE_DECL_ATTRIBUTES (tree @var{olddecl}, tree @var{newdecl})
9470 Define this target hook if the merging of decl attributes needs special
9471 handling. If defined, the result is a list of the combined
9472 @code{DECL_ATTRIBUTES} of @var{olddecl} and @var{newdecl}.
9473 @var{newdecl} is a duplicate declaration of @var{olddecl}. Examples of
9474 when this is needed are when one attribute overrides another, or when an
9475 attribute is nullified by a subsequent definition. This function may
9476 call @code{merge_attributes} to handle machine-independent merging.
9478 @findex TARGET_DLLIMPORT_DECL_ATTRIBUTES
9479 If the only target-specific handling you require is @samp{dllimport}
9480 for Microsoft Windows targets, you should define the macro
9481 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES} to @code{1}. The compiler
9482 will then define a function called
9483 @code{merge_dllimport_decl_attributes} which can then be defined as
9484 the expansion of @code{TARGET_MERGE_DECL_ATTRIBUTES}. You can also
9485 add @code{handle_dll_attribute} in the attribute table for your port
9486 to perform initial processing of the @samp{dllimport} and
9487 @samp{dllexport} attributes. This is done in @file{i386/cygwin.h} and
9488 @file{i386/i386.c}, for example.
9491 @deftypefn {Target Hook} bool TARGET_VALID_DLLIMPORT_ATTRIBUTE_P (tree @var{decl})
9492 @var{decl} is a variable or function with @code{__attribute__((dllimport))}
9493 specified. Use this hook if the target needs to add extra validation
9494 checks to @code{handle_dll_attribute}.
9497 @defmac TARGET_DECLSPEC
9498 Define this macro to a nonzero value if you want to treat
9499 @code{__declspec(X)} as equivalent to @code{__attribute((X))}. By
9500 default, this behavior is enabled only for targets that define
9501 @code{TARGET_DLLIMPORT_DECL_ATTRIBUTES}. The current implementation
9502 of @code{__declspec} is via a built-in macro, but you should not rely
9503 on this implementation detail.
9506 @deftypefn {Target Hook} void TARGET_INSERT_ATTRIBUTES (tree @var{node}, tree *@var{attr_ptr})
9507 Define this target hook if you want to be able to add attributes to a decl
9508 when it is being created. This is normally useful for back ends which
9509 wish to implement a pragma by using the attributes which correspond to
9510 the pragma's effect. The @var{node} argument is the decl which is being
9511 created. The @var{attr_ptr} argument is a pointer to the attribute list
9512 for this decl. The list itself should not be modified, since it may be
9513 shared with other decls, but attributes may be chained on the head of
9514 the list and @code{*@var{attr_ptr}} modified to point to the new
9515 attributes, or a copy of the list may be made if further changes are
9519 @deftypefn {Target Hook} bool TARGET_FUNCTION_ATTRIBUTE_INLINABLE_P (tree @var{fndecl})
9521 This target hook returns @code{true} if it is ok to inline @var{fndecl}
9522 into the current function, despite its having target-specific
9523 attributes, @code{false} otherwise. By default, if a function has a
9524 target specific attribute attached to it, it will not be inlined.
9527 @deftypefn {Target Hook} bool TARGET_VALID_OPTION_ATTRIBUTE_P (tree @var{fndecl}, tree @var{name}, tree @var{args}, int @var{flags})
9528 This hook is called to parse the @code{attribute(option("..."))}, and
9529 it allows the function to set different target machine compile time
9530 options for the current function that might be different than the
9531 options specified on the command line. The hook should return
9532 @code{true} if the options are valid.
9534 The hook should set the @var{DECL_FUNCTION_SPECIFIC_TARGET} field in
9535 the function declaration to hold a pointer to a target specific
9536 @var{struct cl_target_option} structure.
9539 @deftypefn {Target Hook} void TARGET_OPTION_SAVE (struct cl_target_option *@var{ptr})
9540 This hook is called to save any additional target specific information
9541 in the @var{struct cl_target_option} structure for function specific
9543 @xref{Option file format}.
9546 @deftypefn {Target Hook} void TARGET_OPTION_RESTORE (struct cl_target_option *@var{ptr})
9547 This hook is called to restore any additional target specific
9548 information in the @var{struct cl_target_option} structure for
9549 function specific options.
9552 @deftypefn {Target Hook} void TARGET_OPTION_PRINT (struct cl_target_option *@var{ptr})
9553 This hook is called to print any additional target specific
9554 information in the @var{struct cl_target_option} structure for
9555 function specific options.
9558 @deftypefn {Target Hook} bool TARGET_OPTION_PRAGMA_PARSE (target @var{args})
9559 This target hook parses the options for @code{#pragma GCC option} to
9560 set the machine specific options for functions that occur later in the
9561 input stream. The options should be the same as handled by the
9562 @code{TARGET_VALID_OPTION_ATTRIBUTE_P} hook.
9565 @deftypefn {Target Hook} bool TARGET_CAN_INLINE_P (tree @var{caller}, tree @var{callee})
9566 This target hook returns @code{false} if the @var{caller} function
9567 cannot inline @var{callee}, based on target specific information. By
9568 default, inlining is not allowed if the callee function has function
9569 specific target options and the caller does not use the same options.
9573 @section Emulating TLS
9574 @cindex Emulated TLS
9576 For targets whose psABI does not provide Thread Local Storage via
9577 specific relocations and instruction sequences, an emulation layer is
9578 used. A set of target hooks allows this emulation layer to be
9579 configured for the requirements of a particular target. For instance
9580 the psABI may in fact specify TLS support in terms of an emulation
9583 The emulation layer works by creating a control object for every TLS
9584 object. To access the TLS object, a lookup function is provided
9585 which, when given the address of the control object, will return the
9586 address of the current thread's instance of the TLS object.
9588 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_GET_ADDRESS
9589 Contains the name of the helper function that uses a TLS control
9590 object to locate a TLS instance. The default causes libgcc's
9591 emulated TLS helper function to be used.
9594 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_REGISTER_COMMON
9595 Contains the name of the helper function that should be used at
9596 program startup to register TLS objects that are implicitly
9597 initialized to zero. If this is @code{NULL}, all TLS objects will
9598 have explicit initializers. The default causes libgcc's emulated TLS
9599 registration function to be used.
9602 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_SECTION
9603 Contains the name of the section in which TLS control variables should
9604 be placed. The default of @code{NULL} allows these to be placed in
9608 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_SECTION
9609 Contains the name of the section in which TLS initializers should be
9610 placed. The default of @code{NULL} allows these to be placed in any
9614 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_VAR_PREFIX
9615 Contains the prefix to be prepended to TLS control variable names.
9616 The default of @code{NULL} uses a target-specific prefix.
9619 @deftypevr {Target Hook} {const char *} TARGET_EMUTLS_TMPL_PREFIX
9620 Contains the prefix to be prepended to TLS initializer objects. The
9621 default of @code{NULL} uses a target-specific prefix.
9624 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_FIELDS (tree @var{type}, tree *@var{name})
9625 Specifies a function that generates the FIELD_DECLs for a TLS control
9626 object type. @var{type} is the RECORD_TYPE the fields are for and
9627 @var{name} should be filled with the structure tag, if the default of
9628 @code{__emutls_object} is unsuitable. The default creates a type suitable
9629 for libgcc's emulated TLS function.
9632 @deftypefn {Target Hook} tree TARGET_EMUTLS_VAR_INIT (tree @var{var}, tree @var{decl}, tree @var{tmpl_addr})
9633 Specifies a function that generates the CONSTRUCTOR to initialize a
9634 TLS control object. @var{var} is the TLS control object, @var{decl}
9635 is the TLS object and @var{tmpl_addr} is the address of the
9636 initializer. The default initializes libgcc's emulated TLS control object.
9639 @deftypevr {Target Hook} {bool} TARGET_EMUTLS_VAR_ALIGN_FIXED
9640 Specifies whether the alignment of TLS control variable objects is
9641 fixed and should not be increased as some backends may do to optimize
9642 single objects. The default is false.
9645 @deftypevr {Target Hook} {bool} TARGET_EMUTLS_DEBUG_FORM_TLS_ADDRESS
9646 Specifies whether a DWARF @code{DW_OP_form_tls_address} location descriptor
9647 may be used to describe emulated TLS control objects.
9650 @node MIPS Coprocessors
9651 @section Defining coprocessor specifics for MIPS targets.
9652 @cindex MIPS coprocessor-definition macros
9654 The MIPS specification allows MIPS implementations to have as many as 4
9655 coprocessors, each with as many as 32 private registers. GCC supports
9656 accessing these registers and transferring values between the registers
9657 and memory using asm-ized variables. For example:
9660 register unsigned int cp0count asm ("c0r1");
9666 (``c0r1'' is the default name of register 1 in coprocessor 0; alternate
9667 names may be added as described below, or the default names may be
9668 overridden entirely in @code{SUBTARGET_CONDITIONAL_REGISTER_USAGE}.)
9670 Coprocessor registers are assumed to be epilogue-used; sets to them will
9671 be preserved even if it does not appear that the register is used again
9672 later in the function.
9674 Another note: according to the MIPS spec, coprocessor 1 (if present) is
9675 the FPU@. One accesses COP1 registers through standard mips
9676 floating-point support; they are not included in this mechanism.
9678 There is one macro used in defining the MIPS coprocessor interface which
9679 you may want to override in subtargets; it is described below.
9681 @defmac ALL_COP_ADDITIONAL_REGISTER_NAMES
9682 A comma-separated list (with leading comma) of pairs describing the
9683 alternate names of coprocessor registers. The format of each entry should be
9685 @{ @var{alternatename}, @var{register_number}@}
9691 @section Parameters for Precompiled Header Validity Checking
9692 @cindex parameters, precompiled headers
9694 @deftypefn {Target Hook} void *TARGET_GET_PCH_VALIDITY (size_t *@var{sz})
9695 This hook returns the data needed by @code{TARGET_PCH_VALID_P} and sets
9696 @samp{*@var{sz}} to the size of the data in bytes.
9699 @deftypefn {Target Hook} const char *TARGET_PCH_VALID_P (const void *@var{data}, size_t @var{sz})
9700 This hook checks whether the options used to create a PCH file are
9701 compatible with the current settings. It returns @code{NULL}
9702 if so and a suitable error message if not. Error messages will
9703 be presented to the user and must be localized using @samp{_(@var{msg})}.
9705 @var{data} is the data that was returned by @code{TARGET_GET_PCH_VALIDITY}
9706 when the PCH file was created and @var{sz} is the size of that data in bytes.
9707 It's safe to assume that the data was created by the same version of the
9708 compiler, so no format checking is needed.
9710 The default definition of @code{default_pch_valid_p} should be
9711 suitable for most targets.
9714 @deftypefn {Target Hook} const char *TARGET_CHECK_PCH_TARGET_FLAGS (int @var{pch_flags})
9715 If this hook is nonnull, the default implementation of
9716 @code{TARGET_PCH_VALID_P} will use it to check for compatible values
9717 of @code{target_flags}. @var{pch_flags} specifies the value that
9718 @code{target_flags} had when the PCH file was created. The return
9719 value is the same as for @code{TARGET_PCH_VALID_P}.
9723 @section C++ ABI parameters
9724 @cindex parameters, c++ abi
9726 @deftypefn {Target Hook} tree TARGET_CXX_GUARD_TYPE (void)
9727 Define this hook to override the integer type used for guard variables.
9728 These are used to implement one-time construction of static objects. The
9729 default is long_long_integer_type_node.
9732 @deftypefn {Target Hook} bool TARGET_CXX_GUARD_MASK_BIT (void)
9733 This hook determines how guard variables are used. It should return
9734 @code{false} (the default) if first byte should be used. A return value of
9735 @code{true} indicates the least significant bit should be used.
9738 @deftypefn {Target Hook} tree TARGET_CXX_GET_COOKIE_SIZE (tree @var{type})
9739 This hook returns the size of the cookie to use when allocating an array
9740 whose elements have the indicated @var{type}. Assumes that it is already
9741 known that a cookie is needed. The default is
9742 @code{max(sizeof (size_t), alignof(type))}, as defined in section 2.7 of the
9743 IA64/Generic C++ ABI@.
9746 @deftypefn {Target Hook} bool TARGET_CXX_COOKIE_HAS_SIZE (void)
9747 This hook should return @code{true} if the element size should be stored in
9748 array cookies. The default is to return @code{false}.
9751 @deftypefn {Target Hook} int TARGET_CXX_IMPORT_EXPORT_CLASS (tree @var{type}, int @var{import_export})
9752 If defined by a backend this hook allows the decision made to export
9753 class @var{type} to be overruled. Upon entry @var{import_export}
9754 will contain 1 if the class is going to be exported, @minus{}1 if it is going
9755 to be imported and 0 otherwise. This function should return the
9756 modified value and perform any other actions necessary to support the
9757 backend's targeted operating system.
9760 @deftypefn {Target Hook} bool TARGET_CXX_CDTOR_RETURNS_THIS (void)
9761 This hook should return @code{true} if constructors and destructors return
9762 the address of the object created/destroyed. The default is to return
9766 @deftypefn {Target Hook} bool TARGET_CXX_KEY_METHOD_MAY_BE_INLINE (void)
9767 This hook returns true if the key method for a class (i.e., the method
9768 which, if defined in the current translation unit, causes the virtual
9769 table to be emitted) may be an inline function. Under the standard
9770 Itanium C++ ABI the key method may be an inline function so long as
9771 the function is not declared inline in the class definition. Under
9772 some variants of the ABI, an inline function can never be the key
9773 method. The default is to return @code{true}.
9776 @deftypefn {Target Hook} void TARGET_CXX_DETERMINE_CLASS_DATA_VISIBILITY (tree @var{decl})
9777 @var{decl} is a virtual table, virtual table table, typeinfo object,
9778 or other similar implicit class data object that will be emitted with
9779 external linkage in this translation unit. No ELF visibility has been
9780 explicitly specified. If the target needs to specify a visibility
9781 other than that of the containing class, use this hook to set
9782 @code{DECL_VISIBILITY} and @code{DECL_VISIBILITY_SPECIFIED}.
9785 @deftypefn {Target Hook} bool TARGET_CXX_CLASS_DATA_ALWAYS_COMDAT (void)
9786 This hook returns true (the default) if virtual tables and other
9787 similar implicit class data objects are always COMDAT if they have
9788 external linkage. If this hook returns false, then class data for
9789 classes whose virtual table will be emitted in only one translation
9790 unit will not be COMDAT.
9793 @deftypefn {Target Hook} bool TARGET_CXX_LIBRARY_RTTI_COMDAT (void)
9794 This hook returns true (the default) if the RTTI information for
9795 the basic types which is defined in the C++ runtime should always
9796 be COMDAT, false if it should not be COMDAT.
9799 @deftypefn {Target Hook} bool TARGET_CXX_USE_AEABI_ATEXIT (void)
9800 This hook returns true if @code{__aeabi_atexit} (as defined by the ARM EABI)
9801 should be used to register static destructors when @option{-fuse-cxa-atexit}
9802 is in effect. The default is to return false to use @code{__cxa_atexit}.
9805 @deftypefn {Target Hook} bool TARGET_CXX_USE_ATEXIT_FOR_CXA_ATEXIT (void)
9806 This hook returns true if the target @code{atexit} function can be used
9807 in the same manner as @code{__cxa_atexit} to register C++ static
9808 destructors. This requires that @code{atexit}-registered functions in
9809 shared libraries are run in the correct order when the libraries are
9810 unloaded. The default is to return false.
9813 @deftypefn {Target Hook} void TARGET_CXX_ADJUST_CLASS_AT_DEFINITION (tree @var{type})
9814 @var{type} is a C++ class (i.e., RECORD_TYPE or UNION_TYPE) that has just been
9815 defined. Use this hook to make adjustments to the class (eg, tweak
9816 visibility or perform any other required target modifications).
9820 @section Miscellaneous Parameters
9821 @cindex parameters, miscellaneous
9823 @c prevent bad page break with this line
9824 Here are several miscellaneous parameters.
9826 @defmac HAS_LONG_COND_BRANCH
9827 Define this boolean macro to indicate whether or not your architecture
9828 has conditional branches that can span all of memory. It is used in
9829 conjunction with an optimization that partitions hot and cold basic
9830 blocks into separate sections of the executable. If this macro is
9831 set to false, gcc will convert any conditional branches that attempt
9832 to cross between sections into unconditional branches or indirect jumps.
9835 @defmac HAS_LONG_UNCOND_BRANCH
9836 Define this boolean macro to indicate whether or not your architecture
9837 has unconditional branches that can span all of memory. It is used in
9838 conjunction with an optimization that partitions hot and cold basic
9839 blocks into separate sections of the executable. If this macro is
9840 set to false, gcc will convert any unconditional branches that attempt
9841 to cross between sections into indirect jumps.
9844 @defmac CASE_VECTOR_MODE
9845 An alias for a machine mode name. This is the machine mode that
9846 elements of a jump-table should have.
9849 @defmac CASE_VECTOR_SHORTEN_MODE (@var{min_offset}, @var{max_offset}, @var{body})
9850 Optional: return the preferred mode for an @code{addr_diff_vec}
9851 when the minimum and maximum offset are known. If you define this,
9852 it enables extra code in branch shortening to deal with @code{addr_diff_vec}.
9853 To make this work, you also have to define @code{INSN_ALIGN} and
9854 make the alignment for @code{addr_diff_vec} explicit.
9855 The @var{body} argument is provided so that the offset_unsigned and scale
9856 flags can be updated.
9859 @defmac CASE_VECTOR_PC_RELATIVE
9860 Define this macro to be a C expression to indicate when jump-tables
9861 should contain relative addresses. You need not define this macro if
9862 jump-tables never contain relative addresses, or jump-tables should
9863 contain relative addresses only when @option{-fPIC} or @option{-fPIC}
9867 @deftypefn {Target Hook} unsigned int TARGET_CASE_VALUES_THRESHOLD (void)
9868 This function return the smallest number of different values for which it
9869 is best to use a jump-table instead of a tree of conditional branches.
9870 The default is four for machines with a @code{casesi} instruction and
9871 five otherwise. This is best for most machines.
9874 @defmac CASE_USE_BIT_TESTS
9875 Define this macro to be a C expression to indicate whether C switch
9876 statements may be implemented by a sequence of bit tests. This is
9877 advantageous on processors that can efficiently implement left shift
9878 of 1 by the number of bits held in a register, but inappropriate on
9879 targets that would require a loop. By default, this macro returns
9880 @code{true} if the target defines an @code{ashlsi3} pattern, and
9881 @code{false} otherwise.
9884 @defmac WORD_REGISTER_OPERATIONS
9885 Define this macro if operations between registers with integral mode
9886 smaller than a word are always performed on the entire register.
9887 Most RISC machines have this property and most CISC machines do not.
9890 @defmac LOAD_EXTEND_OP (@var{mem_mode})
9891 Define this macro to be a C expression indicating when insns that read
9892 memory in @var{mem_mode}, an integral mode narrower than a word, set the
9893 bits outside of @var{mem_mode} to be either the sign-extension or the
9894 zero-extension of the data read. Return @code{SIGN_EXTEND} for values
9895 of @var{mem_mode} for which the
9896 insn sign-extends, @code{ZERO_EXTEND} for which it zero-extends, and
9897 @code{UNKNOWN} for other modes.
9899 This macro is not called with @var{mem_mode} non-integral or with a width
9900 greater than or equal to @code{BITS_PER_WORD}, so you may return any
9901 value in this case. Do not define this macro if it would always return
9902 @code{UNKNOWN}. On machines where this macro is defined, you will normally
9903 define it as the constant @code{SIGN_EXTEND} or @code{ZERO_EXTEND}.
9905 You may return a non-@code{UNKNOWN} value even if for some hard registers
9906 the sign extension is not performed, if for the @code{REGNO_REG_CLASS}
9907 of these hard registers @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero
9908 when the @var{from} mode is @var{mem_mode} and the @var{to} mode is any
9909 integral mode larger than this but not larger than @code{word_mode}.
9911 You must return @code{UNKNOWN} if for some hard registers that allow this
9912 mode, @code{CANNOT_CHANGE_MODE_CLASS} says that they cannot change to
9913 @code{word_mode}, but that they can change to another integral mode that
9914 is larger then @var{mem_mode} but still smaller than @code{word_mode}.
9917 @defmac SHORT_IMMEDIATES_SIGN_EXTEND
9918 Define this macro if loading short immediate values into registers sign
9922 @defmac FIXUNS_TRUNC_LIKE_FIX_TRUNC
9923 Define this macro if the same instructions that convert a floating
9924 point number to a signed fixed point number also convert validly to an
9928 @deftypefn {Target Hook} int TARGET_MIN_DIVISIONS_FOR_RECIP_MUL (enum machine_mode @var{mode})
9929 When @option{-ffast-math} is in effect, GCC tries to optimize
9930 divisions by the same divisor, by turning them into multiplications by
9931 the reciprocal. This target hook specifies the minimum number of divisions
9932 that should be there for GCC to perform the optimization for a variable
9933 of mode @var{mode}. The default implementation returns 3 if the machine
9934 has an instruction for the division, and 2 if it does not.
9938 The maximum number of bytes that a single instruction can move quickly
9939 between memory and registers or between two memory locations.
9942 @defmac MAX_MOVE_MAX
9943 The maximum number of bytes that a single instruction can move quickly
9944 between memory and registers or between two memory locations. If this
9945 is undefined, the default is @code{MOVE_MAX}. Otherwise, it is the
9946 constant value that is the largest value that @code{MOVE_MAX} can have
9950 @defmac SHIFT_COUNT_TRUNCATED
9951 A C expression that is nonzero if on this machine the number of bits
9952 actually used for the count of a shift operation is equal to the number
9953 of bits needed to represent the size of the object being shifted. When
9954 this macro is nonzero, the compiler will assume that it is safe to omit
9955 a sign-extend, zero-extend, and certain bitwise `and' instructions that
9956 truncates the count of a shift operation. On machines that have
9957 instructions that act on bit-fields at variable positions, which may
9958 include `bit test' instructions, a nonzero @code{SHIFT_COUNT_TRUNCATED}
9959 also enables deletion of truncations of the values that serve as
9960 arguments to bit-field instructions.
9962 If both types of instructions truncate the count (for shifts) and
9963 position (for bit-field operations), or if no variable-position bit-field
9964 instructions exist, you should define this macro.
9966 However, on some machines, such as the 80386 and the 680x0, truncation
9967 only applies to shift operations and not the (real or pretended)
9968 bit-field operations. Define @code{SHIFT_COUNT_TRUNCATED} to be zero on
9969 such machines. Instead, add patterns to the @file{md} file that include
9970 the implied truncation of the shift instructions.
9972 You need not define this macro if it would always have the value of zero.
9975 @anchor{TARGET_SHIFT_TRUNCATION_MASK}
9976 @deftypefn {Target Hook} int TARGET_SHIFT_TRUNCATION_MASK (enum machine_mode @var{mode})
9977 This function describes how the standard shift patterns for @var{mode}
9978 deal with shifts by negative amounts or by more than the width of the mode.
9979 @xref{shift patterns}.
9981 On many machines, the shift patterns will apply a mask @var{m} to the
9982 shift count, meaning that a fixed-width shift of @var{x} by @var{y} is
9983 equivalent to an arbitrary-width shift of @var{x} by @var{y & m}. If
9984 this is true for mode @var{mode}, the function should return @var{m},
9985 otherwise it should return 0. A return value of 0 indicates that no
9986 particular behavior is guaranteed.
9988 Note that, unlike @code{SHIFT_COUNT_TRUNCATED}, this function does
9989 @emph{not} apply to general shift rtxes; it applies only to instructions
9990 that are generated by the named shift patterns.
9992 The default implementation of this function returns
9993 @code{GET_MODE_BITSIZE (@var{mode}) - 1} if @code{SHIFT_COUNT_TRUNCATED}
9994 and 0 otherwise. This definition is always safe, but if
9995 @code{SHIFT_COUNT_TRUNCATED} is false, and some shift patterns
9996 nevertheless truncate the shift count, you may get better code
10000 @defmac TRULY_NOOP_TRUNCATION (@var{outprec}, @var{inprec})
10001 A C expression which is nonzero if on this machine it is safe to
10002 ``convert'' an integer of @var{inprec} bits to one of @var{outprec}
10003 bits (where @var{outprec} is smaller than @var{inprec}) by merely
10004 operating on it as if it had only @var{outprec} bits.
10006 On many machines, this expression can be 1.
10008 @c rearranged this, removed the phrase "it is reported that". this was
10009 @c to fix an overfull hbox. --mew 10feb93
10010 When @code{TRULY_NOOP_TRUNCATION} returns 1 for a pair of sizes for
10011 modes for which @code{MODES_TIEABLE_P} is 0, suboptimal code can result.
10012 If this is the case, making @code{TRULY_NOOP_TRUNCATION} return 0 in
10013 such cases may improve things.
10016 @deftypefn {Target Hook} int TARGET_MODE_REP_EXTENDED (enum machine_mode @var{mode}, enum machine_mode @var{rep_mode})
10017 The representation of an integral mode can be such that the values
10018 are always extended to a wider integral mode. Return
10019 @code{SIGN_EXTEND} if values of @var{mode} are represented in
10020 sign-extended form to @var{rep_mode}. Return @code{UNKNOWN}
10021 otherwise. (Currently, none of the targets use zero-extended
10022 representation this way so unlike @code{LOAD_EXTEND_OP},
10023 @code{TARGET_MODE_REP_EXTENDED} is expected to return either
10024 @code{SIGN_EXTEND} or @code{UNKNOWN}. Also no target extends
10025 @var{mode} to @var{mode_rep} so that @var{mode_rep} is not the next
10026 widest integral mode and currently we take advantage of this fact.)
10028 Similarly to @code{LOAD_EXTEND_OP} you may return a non-@code{UNKNOWN}
10029 value even if the extension is not performed on certain hard registers
10030 as long as for the @code{REGNO_REG_CLASS} of these hard registers
10031 @code{CANNOT_CHANGE_MODE_CLASS} returns nonzero.
10033 Note that @code{TARGET_MODE_REP_EXTENDED} and @code{LOAD_EXTEND_OP}
10034 describe two related properties. If you define
10035 @code{TARGET_MODE_REP_EXTENDED (mode, word_mode)} you probably also want
10036 to define @code{LOAD_EXTEND_OP (mode)} to return the same type of
10039 In order to enforce the representation of @code{mode},
10040 @code{TRULY_NOOP_TRUNCATION} should return false when truncating to
10044 @defmac STORE_FLAG_VALUE
10045 A C expression describing the value returned by a comparison operator
10046 with an integral mode and stored by a store-flag instruction
10047 (@samp{s@var{cond}}) when the condition is true. This description must
10048 apply to @emph{all} the @samp{s@var{cond}} patterns and all the
10049 comparison operators whose results have a @code{MODE_INT} mode.
10051 A value of 1 or @minus{}1 means that the instruction implementing the
10052 comparison operator returns exactly 1 or @minus{}1 when the comparison is true
10053 and 0 when the comparison is false. Otherwise, the value indicates
10054 which bits of the result are guaranteed to be 1 when the comparison is
10055 true. This value is interpreted in the mode of the comparison
10056 operation, which is given by the mode of the first operand in the
10057 @samp{s@var{cond}} pattern. Either the low bit or the sign bit of
10058 @code{STORE_FLAG_VALUE} be on. Presently, only those bits are used by
10061 If @code{STORE_FLAG_VALUE} is neither 1 or @minus{}1, the compiler will
10062 generate code that depends only on the specified bits. It can also
10063 replace comparison operators with equivalent operations if they cause
10064 the required bits to be set, even if the remaining bits are undefined.
10065 For example, on a machine whose comparison operators return an
10066 @code{SImode} value and where @code{STORE_FLAG_VALUE} is defined as
10067 @samp{0x80000000}, saying that just the sign bit is relevant, the
10071 (ne:SI (and:SI @var{x} (const_int @var{power-of-2})) (const_int 0))
10075 can be converted to
10078 (ashift:SI @var{x} (const_int @var{n}))
10082 where @var{n} is the appropriate shift count to move the bit being
10083 tested into the sign bit.
10085 There is no way to describe a machine that always sets the low-order bit
10086 for a true value, but does not guarantee the value of any other bits,
10087 but we do not know of any machine that has such an instruction. If you
10088 are trying to port GCC to such a machine, include an instruction to
10089 perform a logical-and of the result with 1 in the pattern for the
10090 comparison operators and let us know at @email{gcc@@gcc.gnu.org}.
10092 Often, a machine will have multiple instructions that obtain a value
10093 from a comparison (or the condition codes). Here are rules to guide the
10094 choice of value for @code{STORE_FLAG_VALUE}, and hence the instructions
10099 Use the shortest sequence that yields a valid definition for
10100 @code{STORE_FLAG_VALUE}. It is more efficient for the compiler to
10101 ``normalize'' the value (convert it to, e.g., 1 or 0) than for the
10102 comparison operators to do so because there may be opportunities to
10103 combine the normalization with other operations.
10106 For equal-length sequences, use a value of 1 or @minus{}1, with @minus{}1 being
10107 slightly preferred on machines with expensive jumps and 1 preferred on
10111 As a second choice, choose a value of @samp{0x80000001} if instructions
10112 exist that set both the sign and low-order bits but do not define the
10116 Otherwise, use a value of @samp{0x80000000}.
10119 Many machines can produce both the value chosen for
10120 @code{STORE_FLAG_VALUE} and its negation in the same number of
10121 instructions. On those machines, you should also define a pattern for
10122 those cases, e.g., one matching
10125 (set @var{A} (neg:@var{m} (ne:@var{m} @var{B} @var{C})))
10128 Some machines can also perform @code{and} or @code{plus} operations on
10129 condition code values with less instructions than the corresponding
10130 @samp{s@var{cond}} insn followed by @code{and} or @code{plus}. On those
10131 machines, define the appropriate patterns. Use the names @code{incscc}
10132 and @code{decscc}, respectively, for the patterns which perform
10133 @code{plus} or @code{minus} operations on condition code values. See
10134 @file{rs6000.md} for some examples. The GNU Superoptizer can be used to
10135 find such instruction sequences on other machines.
10137 If this macro is not defined, the default value, 1, is used. You need
10138 not define @code{STORE_FLAG_VALUE} if the machine has no store-flag
10139 instructions, or if the value generated by these instructions is 1.
10142 @defmac FLOAT_STORE_FLAG_VALUE (@var{mode})
10143 A C expression that gives a nonzero @code{REAL_VALUE_TYPE} value that is
10144 returned when comparison operators with floating-point results are true.
10145 Define this macro on machines that have comparison operations that return
10146 floating-point values. If there are no such operations, do not define
10150 @defmac VECTOR_STORE_FLAG_VALUE (@var{mode})
10151 A C expression that gives a rtx representing the nonzero true element
10152 for vector comparisons. The returned rtx should be valid for the inner
10153 mode of @var{mode} which is guaranteed to be a vector mode. Define
10154 this macro on machines that have vector comparison operations that
10155 return a vector result. If there are no such operations, do not define
10156 this macro. Typically, this macro is defined as @code{const1_rtx} or
10157 @code{constm1_rtx}. This macro may return @code{NULL_RTX} to prevent
10158 the compiler optimizing such vector comparison operations for the
10162 @defmac CLZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10163 @defmacx CTZ_DEFINED_VALUE_AT_ZERO (@var{mode}, @var{value})
10164 A C expression that indicates whether the architecture defines a value
10165 for @code{clz} or @code{ctz} with a zero operand.
10166 A result of @code{0} indicates the value is undefined.
10167 If the value is defined for only the RTL expression, the macro should
10168 evaluate to @code{1}; if the value applies also to the corresponding optab
10169 entry (which is normally the case if it expands directly into
10170 the corresponding RTL), then the macro should evaluate to @code{2}.
10171 In the cases where the value is defined, @var{value} should be set to
10174 If this macro is not defined, the value of @code{clz} or
10175 @code{ctz} at zero is assumed to be undefined.
10177 This macro must be defined if the target's expansion for @code{ffs}
10178 relies on a particular value to get correct results. Otherwise it
10179 is not necessary, though it may be used to optimize some corner cases, and
10180 to provide a default expansion for the @code{ffs} optab.
10182 Note that regardless of this macro the ``definedness'' of @code{clz}
10183 and @code{ctz} at zero do @emph{not} extend to the builtin functions
10184 visible to the user. Thus one may be free to adjust the value at will
10185 to match the target expansion of these operations without fear of
10190 An alias for the machine mode for pointers. On most machines, define
10191 this to be the integer mode corresponding to the width of a hardware
10192 pointer; @code{SImode} on 32-bit machine or @code{DImode} on 64-bit machines.
10193 On some machines you must define this to be one of the partial integer
10194 modes, such as @code{PSImode}.
10196 The width of @code{Pmode} must be at least as large as the value of
10197 @code{POINTER_SIZE}. If it is not equal, you must define the macro
10198 @code{POINTERS_EXTEND_UNSIGNED} to specify how pointers are extended
10202 @defmac FUNCTION_MODE
10203 An alias for the machine mode used for memory references to functions
10204 being called, in @code{call} RTL expressions. On most CISC machines,
10205 where an instruction can begin at any byte address, this should be
10206 @code{QImode}. On most RISC machines, where all instructions have fixed
10207 size and alignment, this should be a mode with the same size and alignment
10208 as the machine instruction words - typically @code{SImode} or @code{HImode}.
10211 @defmac STDC_0_IN_SYSTEM_HEADERS
10212 In normal operation, the preprocessor expands @code{__STDC__} to the
10213 constant 1, to signify that GCC conforms to ISO Standard C@. On some
10214 hosts, like Solaris, the system compiler uses a different convention,
10215 where @code{__STDC__} is normally 0, but is 1 if the user specifies
10216 strict conformance to the C Standard.
10218 Defining @code{STDC_0_IN_SYSTEM_HEADERS} makes GNU CPP follows the host
10219 convention when processing system header files, but when processing user
10220 files @code{__STDC__} will always expand to 1.
10223 @defmac NO_IMPLICIT_EXTERN_C
10224 Define this macro if the system header files support C++ as well as C@.
10225 This macro inhibits the usual method of using system header files in
10226 C++, which is to pretend that the file's contents are enclosed in
10227 @samp{extern "C" @{@dots{}@}}.
10232 @defmac REGISTER_TARGET_PRAGMAS ()
10233 Define this macro if you want to implement any target-specific pragmas.
10234 If defined, it is a C expression which makes a series of calls to
10235 @code{c_register_pragma} or @code{c_register_pragma_with_expansion}
10236 for each pragma. The macro may also do any
10237 setup required for the pragmas.
10239 The primary reason to define this macro is to provide compatibility with
10240 other compilers for the same target. In general, we discourage
10241 definition of target-specific pragmas for GCC@.
10243 If the pragma can be implemented by attributes then you should consider
10244 defining the target hook @samp{TARGET_INSERT_ATTRIBUTES} as well.
10246 Preprocessor macros that appear on pragma lines are not expanded. All
10247 @samp{#pragma} directives that do not match any registered pragma are
10248 silently ignored, unless the user specifies @option{-Wunknown-pragmas}.
10251 @deftypefun void c_register_pragma (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10252 @deftypefunx void c_register_pragma_with_expansion (const char *@var{space}, const char *@var{name}, void (*@var{callback}) (struct cpp_reader *))
10254 Each call to @code{c_register_pragma} or
10255 @code{c_register_pragma_with_expansion} establishes one pragma. The
10256 @var{callback} routine will be called when the preprocessor encounters a
10260 #pragma [@var{space}] @var{name} @dots{}
10263 @var{space} is the case-sensitive namespace of the pragma, or
10264 @code{NULL} to put the pragma in the global namespace. The callback
10265 routine receives @var{pfile} as its first argument, which can be passed
10266 on to cpplib's functions if necessary. You can lex tokens after the
10267 @var{name} by calling @code{pragma_lex}. Tokens that are not read by the
10268 callback will be silently ignored. The end of the line is indicated by
10269 a token of type @code{CPP_EOF}. Macro expansion occurs on the
10270 arguments of pragmas registered with
10271 @code{c_register_pragma_with_expansion} but not on the arguments of
10272 pragmas registered with @code{c_register_pragma}.
10274 Note that the use of @code{pragma_lex} is specific to the C and C++
10275 compilers. It will not work in the Java or Fortran compilers, or any
10276 other language compilers for that matter. Thus if @code{pragma_lex} is going
10277 to be called from target-specific code, it must only be done so when
10278 building the C and C++ compilers. This can be done by defining the
10279 variables @code{c_target_objs} and @code{cxx_target_objs} in the
10280 target entry in the @file{config.gcc} file. These variables should name
10281 the target-specific, language-specific object file which contains the
10282 code that uses @code{pragma_lex}. Note it will also be necessary to add a
10283 rule to the makefile fragment pointed to by @code{tmake_file} that shows
10284 how to build this object file.
10289 @defmac HANDLE_SYSV_PRAGMA
10290 Define this macro (to a value of 1) if you want the System V style
10291 pragmas @samp{#pragma pack(<n>)} and @samp{#pragma weak <name>
10292 [=<value>]} to be supported by gcc.
10294 The pack pragma specifies the maximum alignment (in bytes) of fields
10295 within a structure, in much the same way as the @samp{__aligned__} and
10296 @samp{__packed__} @code{__attribute__}s do. A pack value of zero resets
10297 the behavior to the default.
10299 A subtlety for Microsoft Visual C/C++ style bit-field packing
10300 (e.g.@: -mms-bitfields) for targets that support it:
10301 When a bit-field is inserted into a packed record, the whole size
10302 of the underlying type is used by one or more same-size adjacent
10303 bit-fields (that is, if its long:3, 32 bits is used in the record,
10304 and any additional adjacent long bit-fields are packed into the same
10305 chunk of 32 bits. However, if the size changes, a new field of that
10306 size is allocated).
10308 If both MS bit-fields and @samp{__attribute__((packed))} are used,
10309 the latter will take precedence. If @samp{__attribute__((packed))} is
10310 used on a single field when MS bit-fields are in use, it will take
10311 precedence for that field, but the alignment of the rest of the structure
10312 may affect its placement.
10314 The weak pragma only works if @code{SUPPORTS_WEAK} and
10315 @code{ASM_WEAKEN_LABEL} are defined. If enabled it allows the creation
10316 of specifically named weak labels, optionally with a value.
10321 @defmac HANDLE_PRAGMA_PACK_PUSH_POP
10322 Define this macro (to a value of 1) if you want to support the Win32
10323 style pragmas @samp{#pragma pack(push[,@var{n}])} and @samp{#pragma
10324 pack(pop)}. The @samp{pack(push,[@var{n}])} pragma specifies the maximum
10325 alignment (in bytes) of fields within a structure, in much the same way as
10326 the @samp{__aligned__} and @samp{__packed__} @code{__attribute__}s do. A
10327 pack value of zero resets the behavior to the default. Successive
10328 invocations of this pragma cause the previous values to be stacked, so
10329 that invocations of @samp{#pragma pack(pop)} will return to the previous
10333 @defmac HANDLE_PRAGMA_PACK_WITH_EXPANSION
10334 Define this macro, as well as
10335 @code{HANDLE_SYSV_PRAGMA}, if macros should be expanded in the
10336 arguments of @samp{#pragma pack}.
10339 @defmac TARGET_DEFAULT_PACK_STRUCT
10340 If your target requires a structure packing default other than 0 (meaning
10341 the machine default), define this macro to the necessary value (in bytes).
10342 This must be a value that would also be valid to use with
10343 @samp{#pragma pack()} (that is, a small power of two).
10348 @defmac HANDLE_PRAGMA_PUSH_POP_MACRO
10349 Define this macro if you want to support the Win32 style pragmas
10350 @samp{#pragma push_macro(macro-name-as-string)} and @samp{#pragma
10351 pop_macro(macro-name-as-string)}. The @samp{#pragma push_macro(
10352 macro-name-as-string)} pragma saves the named macro and via
10353 @samp{#pragma pop_macro(macro-name-as-string)} it will return to the
10358 @defmac DOLLARS_IN_IDENTIFIERS
10359 Define this macro to control use of the character @samp{$} in
10360 identifier names for the C family of languages. 0 means @samp{$} is
10361 not allowed by default; 1 means it is allowed. 1 is the default;
10362 there is no need to define this macro in that case.
10365 @defmac NO_DOLLAR_IN_LABEL
10366 Define this macro if the assembler does not accept the character
10367 @samp{$} in label names. By default constructors and destructors in
10368 G++ have @samp{$} in the identifiers. If this macro is defined,
10369 @samp{.} is used instead.
10372 @defmac NO_DOT_IN_LABEL
10373 Define this macro if the assembler does not accept the character
10374 @samp{.} in label names. By default constructors and destructors in G++
10375 have names that use @samp{.}. If this macro is defined, these names
10376 are rewritten to avoid @samp{.}.
10379 @defmac INSN_SETS_ARE_DELAYED (@var{insn})
10380 Define this macro as a C expression that is nonzero if it is safe for the
10381 delay slot scheduler to place instructions in the delay slot of @var{insn},
10382 even if they appear to use a resource set or clobbered in @var{insn}.
10383 @var{insn} is always a @code{jump_insn} or an @code{insn}; GCC knows that
10384 every @code{call_insn} has this behavior. On machines where some @code{insn}
10385 or @code{jump_insn} is really a function call and hence has this behavior,
10386 you should define this macro.
10388 You need not define this macro if it would always return zero.
10391 @defmac INSN_REFERENCES_ARE_DELAYED (@var{insn})
10392 Define this macro as a C expression that is nonzero if it is safe for the
10393 delay slot scheduler to place instructions in the delay slot of @var{insn},
10394 even if they appear to set or clobber a resource referenced in @var{insn}.
10395 @var{insn} is always a @code{jump_insn} or an @code{insn}. On machines where
10396 some @code{insn} or @code{jump_insn} is really a function call and its operands
10397 are registers whose use is actually in the subroutine it calls, you should
10398 define this macro. Doing so allows the delay slot scheduler to move
10399 instructions which copy arguments into the argument registers into the delay
10400 slot of @var{insn}.
10402 You need not define this macro if it would always return zero.
10405 @defmac MULTIPLE_SYMBOL_SPACES
10406 Define this macro as a C expression that is nonzero if, in some cases,
10407 global symbols from one translation unit may not be bound to undefined
10408 symbols in another translation unit without user intervention. For
10409 instance, under Microsoft Windows symbols must be explicitly imported
10410 from shared libraries (DLLs).
10412 You need not define this macro if it would always evaluate to zero.
10415 @deftypefn {Target Hook} tree TARGET_MD_ASM_CLOBBERS (tree @var{outputs}, tree @var{inputs}, tree @var{clobbers})
10416 This target hook should add to @var{clobbers} @code{STRING_CST} trees for
10417 any hard regs the port wishes to automatically clobber for an asm.
10418 It should return the result of the last @code{tree_cons} used to add a
10419 clobber. The @var{outputs}, @var{inputs} and @var{clobber} lists are the
10420 corresponding parameters to the asm and may be inspected to avoid
10421 clobbering a register that is an input or output of the asm. You can use
10422 @code{tree_overlaps_hard_reg_set}, declared in @file{tree.h}, to test
10423 for overlap with regards to asm-declared registers.
10426 @defmac MATH_LIBRARY
10427 Define this macro as a C string constant for the linker argument to link
10428 in the system math library, or @samp{""} if the target does not have a
10429 separate math library.
10431 You need only define this macro if the default of @samp{"-lm"} is wrong.
10434 @defmac LIBRARY_PATH_ENV
10435 Define this macro as a C string constant for the environment variable that
10436 specifies where the linker should look for libraries.
10438 You need only define this macro if the default of @samp{"LIBRARY_PATH"}
10442 @defmac TARGET_POSIX_IO
10443 Define this macro if the target supports the following POSIX@ file
10444 functions, access, mkdir and file locking with fcntl / F_SETLKW@.
10445 Defining @code{TARGET_POSIX_IO} will enable the test coverage code
10446 to use file locking when exiting a program, which avoids race conditions
10447 if the program has forked. It will also create directories at run-time
10448 for cross-profiling.
10451 @defmac MAX_CONDITIONAL_EXECUTE
10453 A C expression for the maximum number of instructions to execute via
10454 conditional execution instructions instead of a branch. A value of
10455 @code{BRANCH_COST}+1 is the default if the machine does not use cc0, and
10456 1 if it does use cc0.
10459 @defmac IFCVT_MODIFY_TESTS (@var{ce_info}, @var{true_expr}, @var{false_expr})
10460 Used if the target needs to perform machine-dependent modifications on the
10461 conditionals used for turning basic blocks into conditionally executed code.
10462 @var{ce_info} points to a data structure, @code{struct ce_if_block}, which
10463 contains information about the currently processed blocks. @var{true_expr}
10464 and @var{false_expr} are the tests that are used for converting the
10465 then-block and the else-block, respectively. Set either @var{true_expr} or
10466 @var{false_expr} to a null pointer if the tests cannot be converted.
10469 @defmac IFCVT_MODIFY_MULTIPLE_TESTS (@var{ce_info}, @var{bb}, @var{true_expr}, @var{false_expr})
10470 Like @code{IFCVT_MODIFY_TESTS}, but used when converting more complicated
10471 if-statements into conditions combined by @code{and} and @code{or} operations.
10472 @var{bb} contains the basic block that contains the test that is currently
10473 being processed and about to be turned into a condition.
10476 @defmac IFCVT_MODIFY_INSN (@var{ce_info}, @var{pattern}, @var{insn})
10477 A C expression to modify the @var{PATTERN} of an @var{INSN} that is to
10478 be converted to conditional execution format. @var{ce_info} points to
10479 a data structure, @code{struct ce_if_block}, which contains information
10480 about the currently processed blocks.
10483 @defmac IFCVT_MODIFY_FINAL (@var{ce_info})
10484 A C expression to perform any final machine dependent modifications in
10485 converting code to conditional execution. The involved basic blocks
10486 can be found in the @code{struct ce_if_block} structure that is pointed
10487 to by @var{ce_info}.
10490 @defmac IFCVT_MODIFY_CANCEL (@var{ce_info})
10491 A C expression to cancel any machine dependent modifications in
10492 converting code to conditional execution. The involved basic blocks
10493 can be found in the @code{struct ce_if_block} structure that is pointed
10494 to by @var{ce_info}.
10497 @defmac IFCVT_INIT_EXTRA_FIELDS (@var{ce_info})
10498 A C expression to initialize any extra fields in a @code{struct ce_if_block}
10499 structure, which are defined by the @code{IFCVT_EXTRA_FIELDS} macro.
10502 @defmac IFCVT_EXTRA_FIELDS
10503 If defined, it should expand to a set of field declarations that will be
10504 added to the @code{struct ce_if_block} structure. These should be initialized
10505 by the @code{IFCVT_INIT_EXTRA_FIELDS} macro.
10508 @deftypefn {Target Hook} void TARGET_MACHINE_DEPENDENT_REORG ()
10509 If non-null, this hook performs a target-specific pass over the
10510 instruction stream. The compiler will run it at all optimization levels,
10511 just before the point at which it normally does delayed-branch scheduling.
10513 The exact purpose of the hook varies from target to target. Some use
10514 it to do transformations that are necessary for correctness, such as
10515 laying out in-function constant pools or avoiding hardware hazards.
10516 Others use it as an opportunity to do some machine-dependent optimizations.
10518 You need not implement the hook if it has nothing to do. The default
10519 definition is null.
10522 @deftypefn {Target Hook} void TARGET_INIT_BUILTINS ()
10523 Define this hook if you have any machine-specific built-in functions
10524 that need to be defined. It should be a function that performs the
10527 Machine specific built-in functions can be useful to expand special machine
10528 instructions that would otherwise not normally be generated because
10529 they have no equivalent in the source language (for example, SIMD vector
10530 instructions or prefetch instructions).
10532 To create a built-in function, call the function
10533 @code{lang_hooks.builtin_function}
10534 which is defined by the language front end. You can use any type nodes set
10535 up by @code{build_common_tree_nodes} and @code{build_common_tree_nodes_2};
10536 only language front ends that use those two functions will call
10537 @samp{TARGET_INIT_BUILTINS}.
10540 @deftypefn {Target Hook} tree TARGET_BUILTIN_FUNCTION (unsigned @var{code}, bool @var{initialize_p})
10541 Define this hook if you have any machine-specific built-in functions
10542 that need to be defined. It should be a function that returns the
10543 builtin function declaration for the builtin function code @var{code}.
10544 If there is no such builtin and it cannot be initialized at this time
10545 if @var{initialize_p} is true the function should return @code{NULL_TREE}.
10546 If @var{code} is out of range the function should return
10547 @code{error_mark_node}.
10550 @deftypefn {Target Hook} rtx TARGET_EXPAND_BUILTIN (tree @var{exp}, rtx @var{target}, rtx @var{subtarget}, enum machine_mode @var{mode}, int @var{ignore})
10552 Expand a call to a machine specific built-in function that was set up by
10553 @samp{TARGET_INIT_BUILTINS}. @var{exp} is the expression for the
10554 function call; the result should go to @var{target} if that is
10555 convenient, and have mode @var{mode} if that is convenient.
10556 @var{subtarget} may be used as the target for computing one of
10557 @var{exp}'s operands. @var{ignore} is nonzero if the value is to be
10558 ignored. This function should return the result of the call to the
10562 @deftypefn {Target Hook} tree TARGET_RESOLVE_OVERLOADED_BUILTIN (tree @var{fndecl}, tree @var{arglist})
10564 Select a replacement for a machine specific built-in function that
10565 was set up by @samp{TARGET_INIT_BUILTINS}. This is done
10566 @emph{before} regular type checking, and so allows the target to
10567 implement a crude form of function overloading. @var{fndecl} is the
10568 declaration of the built-in function. @var{arglist} is the list of
10569 arguments passed to the built-in function. The result is a
10570 complete expression that implements the operation, usually
10571 another @code{CALL_EXPR}.
10574 @deftypefn {Target Hook} tree TARGET_FOLD_BUILTIN (tree @var{fndecl}, tree @var{arglist}, bool @var{ignore})
10576 Fold a call to a machine specific built-in function that was set up by
10577 @samp{TARGET_INIT_BUILTINS}. @var{fndecl} is the declaration of the
10578 built-in function. @var{arglist} is the list of arguments passed to
10579 the built-in function. The result is another tree containing a
10580 simplified expression for the call's result. If @var{ignore} is true
10581 the value will be ignored.
10584 @deftypefn {Target Hook} const char * TARGET_INVALID_WITHIN_DOLOOP (rtx @var{insn})
10586 Take an instruction in @var{insn} and return NULL if it is valid within a
10587 low-overhead loop, otherwise return a string why doloop could not be applied.
10589 Many targets use special registers for low-overhead looping. For any
10590 instruction that clobbers these this function should return a string indicating
10591 the reason why the doloop could not be applied.
10592 By default, the RTL loop optimizer does not use a present doloop pattern for
10593 loops containing function calls or branch on table instructions.
10596 @defmac MD_CAN_REDIRECT_BRANCH (@var{branch1}, @var{branch2})
10598 Take a branch insn in @var{branch1} and another in @var{branch2}.
10599 Return true if redirecting @var{branch1} to the destination of
10600 @var{branch2} is possible.
10602 On some targets, branches may have a limited range. Optimizing the
10603 filling of delay slots can result in branches being redirected, and this
10604 may in turn cause a branch offset to overflow.
10607 @deftypefn {Target Hook} bool TARGET_COMMUTATIVE_P (rtx @var{x}, @var{outer_code})
10608 This target hook returns @code{true} if @var{x} is considered to be commutative.
10609 Usually, this is just COMMUTATIVE_P (@var{x}), but the HP PA doesn't consider
10610 PLUS to be commutative inside a MEM@. @var{outer_code} is the rtx code
10611 of the enclosing rtl, if known, otherwise it is UNKNOWN.
10614 @deftypefn {Target Hook} rtx TARGET_ALLOCATE_INITIAL_VALUE (rtx @var{hard_reg})
10616 When the initial value of a hard register has been copied in a pseudo
10617 register, it is often not necessary to actually allocate another register
10618 to this pseudo register, because the original hard register or a stack slot
10619 it has been saved into can be used. @code{TARGET_ALLOCATE_INITIAL_VALUE}
10620 is called at the start of register allocation once for each hard register
10621 that had its initial value copied by using
10622 @code{get_func_hard_reg_initial_val} or @code{get_hard_reg_initial_val}.
10623 Possible values are @code{NULL_RTX}, if you don't want
10624 to do any special allocation, a @code{REG} rtx---that would typically be
10625 the hard register itself, if it is known not to be clobbered---or a
10627 If you are returning a @code{MEM}, this is only a hint for the allocator;
10628 it might decide to use another register anyways.
10629 You may use @code{current_function_leaf_function} in the hook, functions
10630 that use @code{REG_N_SETS}, to determine if the hard
10631 register in question will not be clobbered.
10632 The default value of this hook is @code{NULL}, which disables any special
10636 @deftypefn {Target Hook} int TARGET_UNSPEC_MAY_TRAP_P (const_rtx @var{x}, unsigned @var{flags})
10637 This target hook returns nonzero if @var{x}, an @code{unspec} or
10638 @code{unspec_volatile} operation, might cause a trap. Targets can use
10639 this hook to enhance precision of analysis for @code{unspec} and
10640 @code{unspec_volatile} operations. You may call @code{may_trap_p_1}
10641 to analyze inner elements of @var{x} in which case @var{flags} should be
10645 @deftypefn {Target Hook} void TARGET_SET_CURRENT_FUNCTION (tree @var{decl})
10646 The compiler invokes this hook whenever it changes its current function
10647 context (@code{cfun}). You can define this function if
10648 the back end needs to perform any initialization or reset actions on a
10649 per-function basis. For example, it may be used to implement function
10650 attributes that affect register usage or code generation patterns.
10651 The argument @var{decl} is the declaration for the new function context,
10652 and may be null to indicate that the compiler has left a function context
10653 and is returning to processing at the top level.
10654 The default hook function does nothing.
10656 GCC sets @code{cfun} to a dummy function context during initialization of
10657 some parts of the back end. The hook function is not invoked in this
10658 situation; you need not worry about the hook being invoked recursively,
10659 or when the back end is in a partially-initialized state.
10662 @defmac TARGET_OBJECT_SUFFIX
10663 Define this macro to be a C string representing the suffix for object
10664 files on your target machine. If you do not define this macro, GCC will
10665 use @samp{.o} as the suffix for object files.
10668 @defmac TARGET_EXECUTABLE_SUFFIX
10669 Define this macro to be a C string representing the suffix to be
10670 automatically added to executable files on your target machine. If you
10671 do not define this macro, GCC will use the null string as the suffix for
10675 @defmac COLLECT_EXPORT_LIST
10676 If defined, @code{collect2} will scan the individual object files
10677 specified on its command line and create an export list for the linker.
10678 Define this macro for systems like AIX, where the linker discards
10679 object files that are not referenced from @code{main} and uses export
10683 @defmac MODIFY_JNI_METHOD_CALL (@var{mdecl})
10684 Define this macro to a C expression representing a variant of the
10685 method call @var{mdecl}, if Java Native Interface (JNI) methods
10686 must be invoked differently from other methods on your target.
10687 For example, on 32-bit Microsoft Windows, JNI methods must be invoked using
10688 the @code{stdcall} calling convention and this macro is then
10689 defined as this expression:
10692 build_type_attribute_variant (@var{mdecl},
10694 (get_identifier ("stdcall"),
10699 @deftypefn {Target Hook} bool TARGET_CANNOT_MODIFY_JUMPS_P (void)
10700 This target hook returns @code{true} past the point in which new jump
10701 instructions could be created. On machines that require a register for
10702 every jump such as the SHmedia ISA of SH5, this point would typically be
10703 reload, so this target hook should be defined to a function such as:
10707 cannot_modify_jumps_past_reload_p ()
10709 return (reload_completed || reload_in_progress);
10714 @deftypefn {Target Hook} int TARGET_BRANCH_TARGET_REGISTER_CLASS (void)
10715 This target hook returns a register class for which branch target register
10716 optimizations should be applied. All registers in this class should be
10717 usable interchangeably. After reload, registers in this class will be
10718 re-allocated and loads will be hoisted out of loops and be subjected
10719 to inter-block scheduling.
10722 @deftypefn {Target Hook} bool TARGET_BRANCH_TARGET_REGISTER_CALLEE_SAVED (bool @var{after_prologue_epilogue_gen})
10723 Branch target register optimization will by default exclude callee-saved
10725 that are not already live during the current function; if this target hook
10726 returns true, they will be included. The target code must than make sure
10727 that all target registers in the class returned by
10728 @samp{TARGET_BRANCH_TARGET_REGISTER_CLASS} that might need saving are
10729 saved. @var{after_prologue_epilogue_gen} indicates if prologues and
10730 epilogues have already been generated. Note, even if you only return
10731 true when @var{after_prologue_epilogue_gen} is false, you still are likely
10732 to have to make special provisions in @code{INITIAL_ELIMINATION_OFFSET}
10733 to reserve space for caller-saved target registers.
10736 @defmac POWI_MAX_MULTS
10737 If defined, this macro is interpreted as a signed integer C expression
10738 that specifies the maximum number of floating point multiplications
10739 that should be emitted when expanding exponentiation by an integer
10740 constant inline. When this value is defined, exponentiation requiring
10741 more than this number of multiplications is implemented by calling the
10742 system library's @code{pow}, @code{powf} or @code{powl} routines.
10743 The default value places no upper bound on the multiplication count.
10746 @deftypefn Macro void TARGET_EXTRA_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10747 This target hook should register any extra include files for the
10748 target. The parameter @var{stdinc} indicates if normal include files
10749 are present. The parameter @var{sysroot} is the system root directory.
10750 The parameter @var{iprefix} is the prefix for the gcc directory.
10753 @deftypefn Macro void TARGET_EXTRA_PRE_INCLUDES (const char *@var{sysroot}, const char *@var{iprefix}, int @var{stdinc})
10754 This target hook should register any extra include files for the
10755 target before any standard headers. The parameter @var{stdinc}
10756 indicates if normal include files are present. The parameter
10757 @var{sysroot} is the system root directory. The parameter
10758 @var{iprefix} is the prefix for the gcc directory.
10761 @deftypefn Macro void TARGET_OPTF (char *@var{path})
10762 This target hook should register special include paths for the target.
10763 The parameter @var{path} is the include to register. On Darwin
10764 systems, this is used for Framework includes, which have semantics
10765 that are different from @option{-I}.
10768 @deftypefn {Target Hook} bool TARGET_USE_LOCAL_THUNK_ALIAS_P (tree @var{fndecl})
10769 This target hook returns @code{true} if it is safe to use a local alias
10770 for a virtual function @var{fndecl} when constructing thunks,
10771 @code{false} otherwise. By default, the hook returns @code{true} for all
10772 functions, if a target supports aliases (i.e.@: defines
10773 @code{ASM_OUTPUT_DEF}), @code{false} otherwise,
10776 @defmac TARGET_FORMAT_TYPES
10777 If defined, this macro is the name of a global variable containing
10778 target-specific format checking information for the @option{-Wformat}
10779 option. The default is to have no target-specific format checks.
10782 @defmac TARGET_N_FORMAT_TYPES
10783 If defined, this macro is the number of entries in
10784 @code{TARGET_FORMAT_TYPES}.
10787 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES
10788 If defined, this macro is the name of a global variable containing
10789 target-specific format overrides for the @option{-Wformat} option. The
10790 default is to have no target-specific format overrides. If defined,
10791 @code{TARGET_FORMAT_TYPES} must be defined, too.
10794 @defmac TARGET_OVERRIDES_FORMAT_ATTRIBUTES_COUNT
10795 If defined, this macro specifies the number of entries in
10796 @code{TARGET_OVERRIDES_FORMAT_ATTRIBUTES}.
10799 @defmac TARGET_OVERRIDES_FORMAT_INIT
10800 If defined, this macro specifies the optional initialization
10801 routine for target specific customizations of the system printf
10802 and scanf formatter settings.
10805 @deftypefn {Target Hook} bool TARGET_RELAXED_ORDERING
10806 If set to @code{true}, means that the target's memory model does not
10807 guarantee that loads which do not depend on one another will access
10808 main memory in the order of the instruction stream; if ordering is
10809 important, an explicit memory barrier must be used. This is true of
10810 many recent processors which implement a policy of ``relaxed,''
10811 ``weak,'' or ``release'' memory consistency, such as Alpha, PowerPC,
10812 and ia64. The default is @code{false}.
10815 @deftypefn {Target Hook} const char *TARGET_INVALID_ARG_FOR_UNPROTOTYPED_FN (tree @var{typelist}, tree @var{funcdecl}, tree @var{val})
10816 If defined, this macro returns the diagnostic message when it is
10817 illegal to pass argument @var{val} to function @var{funcdecl}
10818 with prototype @var{typelist}.
10821 @deftypefn {Target Hook} {const char *} TARGET_INVALID_CONVERSION (tree @var{fromtype}, tree @var{totype})
10822 If defined, this macro returns the diagnostic message when it is
10823 invalid to convert from @var{fromtype} to @var{totype}, or @code{NULL}
10824 if validity should be determined by the front end.
10827 @deftypefn {Target Hook} {const char *} TARGET_INVALID_UNARY_OP (int @var{op}, tree @var{type})
10828 If defined, this macro returns the diagnostic message when it is
10829 invalid to apply operation @var{op} (where unary plus is denoted by
10830 @code{CONVERT_EXPR}) to an operand of type @var{type}, or @code{NULL}
10831 if validity should be determined by the front end.
10834 @deftypefn {Target Hook} {const char *} TARGET_INVALID_BINARY_OP (int @var{op}, tree @var{type1}, tree @var{type2})
10835 If defined, this macro returns the diagnostic message when it is
10836 invalid to apply operation @var{op} to operands of types @var{type1}
10837 and @var{type2}, or @code{NULL} if validity should be determined by
10841 @deftypefn {Target Hook} {const char *} TARGET_INVALID_PARAMETER_TYPE (tree @var{type})
10842 If defined, this macro returns the diagnostic message when it is
10843 invalid for functions to include parameters of type @var{type},
10844 or @code{NULL} if validity should be determined by
10845 the front end. This is currently used only by the C and C++ front ends.
10848 @deftypefn {Target Hook} {const char *} TARGET_INVALID_RETURN_TYPE (tree @var{type})
10849 If defined, this macro returns the diagnostic message when it is
10850 invalid for functions to have return type @var{type},
10851 or @code{NULL} if validity should be determined by
10852 the front end. This is currently used only by the C and C++ front ends.
10855 @deftypefn {Target Hook} {tree} TARGET_PROMOTED_TYPE (tree @var{type})
10856 If defined, this target hook returns the type to which values of
10857 @var{type} should be promoted when they appear in expressions,
10858 analogous to the integer promotions, or @code{NULL_TREE} to use the
10859 front end's normal promotion rules. This hook is useful when there are
10860 target-specific types with special promotion rules.
10861 This is currently used only by the C and C++ front ends.
10864 @deftypefn {Target Hook} {tree} TARGET_CONVERT_TO_TYPE (tree @var{type}, tree @var{expr})
10865 If defined, this hook returns the result of converting @var{expr} to
10866 @var{type}. It should return the converted expression,
10867 or @code{NULL_TREE} to apply the front end's normal conversion rules.
10868 This hook is useful when there are target-specific types with special
10870 This is currently used only by the C and C++ front ends.
10873 @defmac TARGET_USE_JCR_SECTION
10874 This macro determines whether to use the JCR section to register Java
10875 classes. By default, TARGET_USE_JCR_SECTION is defined to 1 if both
10876 SUPPORTS_WEAK and TARGET_HAVE_NAMED_SECTIONS are true, else 0.
10880 This macro determines the size of the objective C jump buffer for the
10881 NeXT runtime. By default, OBJC_JBLEN is defined to an innocuous value.
10884 @defmac LIBGCC2_UNWIND_ATTRIBUTE
10885 Define this macro if any target-specific attributes need to be attached
10886 to the functions in @file{libgcc} that provide low-level support for
10887 call stack unwinding. It is used in declarations in @file{unwind-generic.h}
10888 and the associated definitions of those functions.
10891 @deftypefn {Target Hook} void TARGET_UPDATE_STACK_BOUNDARY (void)
10892 Define this macro to update the current function stack boundary if
10896 @deftypefn {Target Hook} rtx TARGET_GET_DRAP_RTX (void)
10897 Define this macro to an rtx for Dynamic Realign Argument Pointer if a
10898 different argument pointer register is needed to access the function's
10899 argument list when stack is aligned.
10902 @deftypefn {Target Hook} {bool} TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS (void)
10903 When optimization is disabled, this hook indicates whether or not
10904 arguments should be allocated to stack slots. Normally, GCC allocates
10905 stacks slots for arguments when not optimizing in order to make
10906 debugging easier. However, when a function is declared with
10907 @code{__attribute__((naked))}, there is no stack frame, and the compiler
10908 cannot safely move arguments from the registers in which they are passed
10909 to the stack. Therefore, this hook should return true in general, but
10910 false for naked functions. The default implementation always returns true.
10914 @deftypevr {Target Hook} {unsigned HOST_WIDE_INT} TARGET_CONST_ANCHOR
10915 On some architectures it can take multiple instructions to synthesize
10916 a constant. If there is another constant already in a register that
10917 is close enough in value then it is preferable that the new constant
10918 is computed from this register using immediate addition or
10919 substraction. We accomplish this through CSE. Besides the value of
10920 the constant we also add a lower and an upper constant anchor to the
10921 available expressions. These are then queried when encountering new
10922 constants. The anchors are computed by rounding the constant up and
10923 down to a multiple of the value of @code{TARGET_CONST_ANCHOR}.
10924 @code{TARGET_CONST_ANCHOR} should be the maximum positive value
10925 accepted by immediate-add plus one. We currently assume that the
10926 value of @code{TARGET_CONST_ANCHOR} is a power of 2. For example, on
10927 MIPS, where add-immediate takes a 16-bit signed value,
10928 @code{TARGET_CONST_ANCHOR} is set to @samp{0x8000}. The default value
10929 is zero, which disables this optimization. @end deftypevr